Weld (Theory of Errors and Least Squares) 1916

GIFT
OF
the estate of
Professor William F. Meyer
(3*^
THEORY OF ERRORS AND

LEAST SQUARES
THE MACMILLAN COMPANY

NEW YORK
BOSTON
CHICAGO DALLAS
SAN FRANCISCO
ATLANTA
MACMILLAN &
LONDON
CO., LIMITED
BOMBAY CALCUTTA
MELBOURNE
THE MACMILLAN
CO. OF
TORONTO
CANADA,
LTD.

LEAST SQUARES
A TEXTBOOK FOR COLLEGE STUDENTS
AND RESEARCH WORKERS
BY
LsROY
D.
WELD,
M.S.
PROFESSOR OF PHYSICS IN COE COLLEGE
gorfc

1916
All rights reserved
COPYRIGHT, 1916,
BY THE MACMILLAN COMPANY.

Set
up and electrotyped.
J. 8.
Published March, 1916.
Berwick & Smith Co.

Gushing Co.
Norwood, Mass., U.S.A.
PREFACE
THERE
are
few branches
of mathematics
which have
wider applicability to general scientific work than the

Theory of Errors, and few mathematical implements
which are capable of greater usefulness to the research

worker than the Method of Least Squares. Yet, for
some reason, students are rarely given opportunity
to
acquire facility in these lines, the result being that too
many
and engineers go about their

It would be almost
enumerate the variety of ways in which
of our scientists
work without such equipment.
impossible to
the ideas relating to these subjects adapt themselves
to even such simple bits of quantitative work as the
chemist or the surveyor is daily called upon to do.

And it is difficult for the writer to imagine how an
elaborate research in any of the exact sciences can
be carried on at
all, without the constant application

these principles throughout both the preliminary
and the final stages of the work. The satisfaction to
of
be gained from the application of the theory of pre

cision alone is well worth all the time necessary to
acquire these subjects.
Add
to this the fact that the
M577305
PREFACE
VI
theory of error distribution has direct theoretical bear

ing upon certain very important laws and problems of
physics, chemistry, astronomy, and even of biology,
and the reasons for students having opportunity to

attain the elements of the subject become still more
emphatic.
This small volume embodies the material used by
the writer as lecture notes during the past twelve
intended as a presentation of the Theory

of Errors and Least Squares in such a simple and con
years.
It is
form as to be useful, not only as a textbook for

undergraduates, but as a handy reference which any
cise
research worker can read through in an evening or so
and then put into immediate
practice.
be noticed that the illustrative examples and

problems are drawn from various branches of science,
It will
No
suggesting the wide range of possible application.
is
of
at
an
exhaustive
treatment
made,
course,
attempt
in such small compass.
Some
of the special
methods
employed by expert computers, often included in larger

works, have been purposely omitted. For the conven
ience of the student, and in order not to interrupt the
thread of the subject, a few of the more complicated

mathematical discussions have been set apart in the
Appendix and referred
to at the appropriate places.
It is not intended that they shall be omitted
from the
when using the book as a text, though the cas

ual reader may get along very well without them.
The writer wishes to express his appreciation to the
numerous friends who have kindly given aid by way
course
PREFACE
Vll
data for the illustrative examples, or

Where material has been taken from other
of furnishing
otherwise.
works, due credit has been given for the same.

L. D.
CEDAR RAPIDS, IOWA,

December, 1915.
W.
CONTENTS
CHAPTER
ON MEASUREMENT
PAGE
ARTICLE
Definition of
2.
measurement
Indirect measurement
3.
Estimation
1.
4.
The
5.
Errors of measurement
6.
Exercises
...
.......
impossibility of exact measurements
5
6
CHAPTER
II
ON THE OCCURRENCE AND GENERAL

PROPERTIES OF ERRORS
7.
Errors and residuals
8.
Classification of errors
9.
Mistakes
General methods of eliminating persistent errors
Exercises leading to an understanding of error
bution
10.
11.
13.
Precision
14.
Mathematical expression of the law of error
distribution of errors
.
CHAPTER
16.
17.
.
-.
.
principle
Definition of mathematical probability
.
ix
18
distri
.
28
30
III
Fundamental
Permutations
17
.
26
ON PROBABILITIES
15.
.11
.22
....
....
...
Remarks on the
13
12.
....32
31
-*
34
CONTENTS
X
ARTICLE
18.
PAGE
Combinations
.36
...
.
Probability of either of two or more events

Probability of the concurrence of independent events
21. The coin problem
19.
........
20.
22.
23.
4.
Important exercise
Empirical or statistical probability
39
40
41
44
...
Exercises
45
45
CHAPTER IV
THE ERROR EQUATION AND THE PRINCIPLE

OF LEAST SQUARES
25.
26.
.
Analogy of error distribution to coin problem
The most probable value from a series of direct measure
The
ments.
27.
Gauss
29.
The
30.
Exercises
arithmetical
mean
....
.....
51
simplest form
58
deduction of the error equation

28. Discussion of the error equation
s
principle of least squares in

.
its
49
52
56
61
CHAPTER V
ON THE ADJUSTMENT OF INDIRECT

OBSERVATIONS
Observations on functions of a single quantity
32. Observation equations for more than one
31.
quantity
33.
More observations than
34.
Reduction of observation equations of the
35.
Illustrations
36.
Illustrations
37.
Illustrations
38.
Illustrations
from
from
from
from
first
degree
Exercises
69
72
74
....
surveying
.81
astronomy
78
chemistry
...
.
ditions
41.
physics
Observation equations not of first degree

.
40. Observations upon quantities subject to rigorous con
39.
65
67
Normal equations
quantities.
unknown
85
.89
91
94
CONTENTS
XI
CHAPTER VI
EMPIRICAL FORMULAS
PAGE
AUTICLB
42.
Classification of formulas
.....
.....
Uses and limitations of empirical formulas

44. Illustrations of empirical formulas
45. Choice of mathematical expression
43.
46.
104
105
107
Ill
115
Exercises
CHAPTER
VII
WEIGHTED OBSERVATIONS
48.
Relative reliability of observations. Weights

Adjustment of observations of unequal weight
49.
Exercises
50.
51.
Wild or doubtful observations

The precision index h
52.
General statement of the principle of least squares
47.
...
.
CHAPTER
124
126
128
.
.*,
.
135
136
139
VIII
PRECISION AND THE PROBABLE ERROR

54.
Discontinuity of the error variable

Value of the integral /, and the relation between c and k
55.
Probability of an error lying between given limits.
56.
probability integral
Calculation of the precision index from the residuals
57.
Approximate formulas for the precision index
58.
The probable
53.
.......
error of an observation
Relation between probable error and weight

60. Exercises
61. Probable errors of functions of observed quantities
.
63.
Probable errors of adjusted values

Probable errors of conditioned observations
64.
Exercises
62.
143
The
.
....
59.
141
144
146
149
155
163
152
159
166
.
170
.171
CONTENTS
Xll
APPENDIX
SUPPLEMENTARY NOTES
PAGE
A. Proof of the necessary functional relation assumed in

deriving the error law. (Supplementary to Art. 27)
B. Approximation method for observation equations not of
the
first
(Supplementary to Art. 39)
degree.
Evaluation of the integral
D.
to Art. 54)
Evaluation of the probability integral.
to Art. 55)
.
178
/*
C.
177
e-tf^dx.
(Supplementary
180
(Supplementary
182
E. Outline of another method for probable errors of adjusted

values.
(Supplementary to Art. 62)
...
183
F. Collection of important definitions, theorems, rules, and

formulas for convenient reference
185

LEAST SQUARES
THEORY OF ERRORS AND LEAST

SQUARES
CHAPTER
ON MEASUREMENT
Definition of
1.
tity
ratio to the unit
by any
employed
in
Thus,
quantity.
To
Measurement.
to determine
is
measure a quan
means, direct or indirect, its
in expressing the value of that
measuring a
line,
we
find that
it
is
a certain number of times as long as the foot or the centi

meter, and this
number
is
said to be its value in feet or
centimeters.
This definition must be clearly understood to be in

dependent of whatever process is used in the measurement.
We
could measure the area of a polygonal piece of sheet
two ways either by measuring its sides and angles

and computing its area by geometry, or by weighing it
and comparing its weight to that of a square piece with
iron in
unit side.
ment
2.
Either of these processes
of the area,
Indirect
though neither
Measurement.
is
is
a true measure
a direct measurement.
Indeed, with the excep
tion of one kind of magnitude, very few

are direct.
B
By
this
is
meant that we do
1
measurements
not, in general,
THEORY OF ERRORS AND LEAST SQUARES
apply the unit of measure directly to the magnitude to

This is done commonly only in the case of
be measured.
length.
We can,
in
measuring a
line,
apply the yardstick
directly along the line and determine in this
many times
greater one
is
than the other.
way how
But we cannot
take a lamp in one hand and a standard candle in the

other and determine the candle-power of the lamp in
any such direct manner.

So far is the above mentioned
matter of
fact,
made to depend,
This
will
principle true, that, as a
nearly every kind of measurement

in practice,
upon measurements
be clear from a number of
is
of length.
illustrations.
Angles are measured, not by applying the wedge-like

degree as a unit, but by measuring the length of the arc
laid off
on a curved
linear scale, or
by measuring the
and com
lengths of straight lines connected with the angle
puting the latter from
its
trigonometric functions.
measured, not by counting the minutes and

seconds in the interval, but by observing the motion of
Time
is
the clock hands over a curvilinear scale called the dial,
marked
off in
spaces of equal length;
or
by noting the
marked off on the chronograph record by a pen

%T
is p- en a lateral jerk electrically at the
which
point
beginning and end of the interval. Every magnitude
measured off on a dial is finally referred to length, as
lengths
exemplified
by pressure gauges, gas meters,
electric
me
ters, aneroid barometers, etc.
Temperature
is
the thermometer.
measured
off as
a length on the stem of
ON MEASUREMENT
Atmospheric pressure
is
measured, and even expressed,
in inches or centimeters of mercury.
Weight
measured, in the
is
on a
position of a slide or rider
work by the
final
adjustment, by the
linear scale, or in refined
position of the balance pointer at equilibrium,
the sensibility of the balance being known.

spring balance and
its
more
The common
refined near relative, the Jolly
balance, illustrate the linear principle in
another way.
In short, every measuring instrument has some sort of

on which some sort
linear scale, either straight or curved,
of indicator or pointer
The reason
moves.
for thus referring every kind of
to a simple one of length

to,
that length
is
is
measurement
mainly the one already referred
the only kind of magnitude that can
be conveniently compared directly with its own unit.

But there is another reason. The eye can estimate a
length with far greater accuracy than the muscles can
estimate a weight, the hand a temperature, or the con

sciousness an interval of time;
and
this process of esti
mation plays an all-important part, as will now be seen,

in every kind of accurate measurement.
3.
Estimation.
The degree
of precision
with which
an observer can read a given linear scale depends upon

two things, namely: (1) the definiteness or sharpness of
the marks on the scale and of the pointer or indicator,
and
(2)
the
skill
with which the observer can estimate
fractional parts of one interval or scale-division.
The former item may be made
clear
by comparing the
scale
and indicator of an ordinary spring balance with those
on a delicate ampere meter or aneroid

the graduations on a surveyor
or
barometer;
leveling-rod with those on
a silver-inlaid standard meter bar.
As to the second matter,

that the observer
drill
it is
of the
utmost importance
himself in this process of estimation.
In no case does the accuracy of a single scale-reading end
with the fineness of graduation of the
scale,
providing
the scale lines and indicator are sharp and distinct;
it
can always be carried a step farther.

It is the custom of practical observers to make estima
tions of fractional units in tenths, not in halves, thirds,
etc.,
is
and to record the readings decimally.
made
to estimate the hundredths, unless
the observer that the fraction

three-fourths,
.75
even this
linear scale
is
may
attempt
appears to
exactly one-fourth or
is
when he would be
No
it
likely to record .25 or
a doubtful practice.
The reading
of
any
be carried, in general, to an accuracy of
one-tenth of the smallest scale-division by the estimation

of the eye alone, or the eye aided
by a magnifier
if
desir
able.
In
many
instruments of precision, the linear scale
is
provided with some sort of vernier, which is a mechanical

substitute for the estimation of fractional parts of scale
divisions.
in use
may
Descriptions of the different kinds of verniers
be found in any elementary laboratory manual
of physics, or in
any encyclopedia.
the vernier requires the same sort of
But even the use

skill
of
and judgment as
estimation, namely, a correct idea of linear position and
ON MEASUREMENT
And
coincidence.
in
the vast majority of measuring
instruments, no vernier
provided, and the observer
is
must be able to estimate tenths accurately and without

hesitation.
4.
The
scientist
Impossibility of Exact
is
familiar
with
Measurements.
that
the fact
there
Every
is no
such thing as an absolutely exact measurement, for the

simple reason- that the quantity measured and the unit
measure are never commensurable.
of
If
we weigh
carefully a small piece of metal
on a common
balance, a typical result would be 3.9843 grams, and not a

is, however, only an
even
if correct to four
approximation to the true weight,
decimal places, just as the number 3.1416 is only an ap
whole number, as four grams.
This
proximation to the value of IT. If a more sensitive bal

ance is used, the result may be 3.984326 grams but as the
;
masses of the piece of metal and the
gram weight
commensurable, the true weight, even if

to weigh without the inaccuracies that
it
are in
were possible
from im
arise
and judgment in estimation, would be

grams, and the result obtained could
perfect apparatus
inexpressible
in
be true only to the degree
by
six places of decimals,
of a
of
approximation represented
that
is,
to the nearest millionth
gram.
What
it will
is
true of weighing
is
true of
all
measurement, and
readily be seen that to obtain the true value of
actual concrete quantity
true value of V^, or
TT,
is
any
as hopeless as to obtain the
or logio 17.
Errors of Measurement.
5.
incommensurability
more
far
Aside from the mere
of magnitudes, there is another
and
serious hindrance to the obtaining of correct
values
by measurement, and
known
as error.
Suppose the
bit of metal,
sensitive balance to
this is
what
is
technically
which was found on the more
weigh 3.984326 grams, be now weighed
by the same person, in the same room, on the same

balance and with the same weights. More likely than
again,
not, the result will turn out to be different
from the former
by some millionths of a gram, perhaps thirty or

This means simply that neither re
forty millionths.
result
even to the sixth decimal place.

we go out with a surveyor s transit of the
sult is correct,
Again,
if
finest
construction and measure with the utmost care, to seconds

even, each of the three angles of a triangle
marked out
accurately centered stakes on level ground, and add
by
we shall probably find that

from 180 by several seconds one way or
may repeat the operation with equal
the three results together,

their
sum
differs
We
the other.
care
and
skill,
and get a
still
than the
first.
farther from 180

all
different result, perhaps
This illustration
will
be
the more striking, in that in this case the true value
of the
sum
of the three angles is
known from geometry,
while in the case of the weights the true value
and can never be known.
Even
is
not
here, the individual
angles cannot be obtained exactly.
The
and
causes of error in precise measurements are
various.
single
example
many
will suffice to illustrate
ON MEASUREMENT
Suppose we wish to measure the distance from one

stake to another with a surveyor s chain. Two men
this.
carry the chain.
Each time they advance, one
adjusts the
following end to the rear marking-pin, the other sets a
new pin
at the leading end,
with absolute accuracy.
and neither can do
They
this
work
do not stretch the chain
tight enough; they do not hold the chain horizontal in

going up or down hill they do not follow a straight line
;
they do not notice kinks in the chain, and they neglect

the fact that the chain
As a consequence
longer.
many
wearing at the joints and getting
is
of all these small items,
others not mentioned, the measurement
the end be several inches from the truth
if
and
may
in
the line to be
measured be very long. This is only one instance showing

how hundreds of little disturbances may combine and
form one
which
be positive or
negative, great or small, according to which kind of dis
turbances predominates (that is, whether they tend to
make the result too large or too small), and to whether
final resultant error
may
they happen to be about evenly balanced or not.
A systematic study of
the occurrence of errors gives rise
to a mathematical analysis, based essentially upon the

principles of probability
and known as the Theory
of
Errors; and our attempts to apply this theory to the re
measurements, with a view to getting the values

that are probably nearest the truth, have resulted in
sults of
the formulation of certain rules embraced in that part of

the error theory
known
as the Method of Least Squares.
EXERCISES
6.
The
following exercises are intended for the use
of students
who have not done much
had the advantage

or field work.
gestive,
laboratory work nor
of a course in laboratory
It will
measurements
be seen that they are largely sug

be modified as desired to suit the
and they may

For advanced students and research
circumstances.
workers they
be omitted altogether.
may
1.
Can you think of any kind of accurate measure
ment not ultimately employing some sort of linear
scale ?
Show wherein
made to employ
2.
are
density of a solid
the following kinds of measurements
a linear scale
relative
area of a piece of land
humidity
of the
atmosphere
index of refraction of a transparent substance

liquid
3.
volume
of
from a burette.
Determine the volume
of a material sphere, cylinder
two ways first by measuring

and
second
dimensions;
by dropping it into a glass
graduate partly filled with water and observing the
or other geometrical solid in
its
displacement.
consider the
Do
more
the two results agree
precise
Which do you
method ?
pure water in two ways

and second by
first by placing it in a glass graduate;
balance
and
it
on
a
computing the volume.
weighing
the
done
in
The weighing may be
graduate, which has
4.
Measure a quantity
been weighed beforehand.
of
ON MEASUREMENT
Lay
5.
on a sheet
off
of
smooth paper, with a fine, hard

mark two points on
pencil, a line of indefinite length, and

it
at
random somewhat
straight edge
less
apart as possible.
On
than 10 cm. apart.
mark two
By means of
of a card,
the
points as nearly 10 cm.

direct comparison with
this standard, estimate the length of the first line-segment
down
in centimeters, writing
the
unknown
line
Next compare
the result.
with a cardboard scale marked
centimeters but not in millimeters, observing the

ber of centimeters and estimating the millimeters.
compare the same
line
with a millimeter
the tenths of a millimeter.
Notice
how
scale,
off in
num
Finally
estimating
the three results
agree, all being expressed in centimeters.
Repeat
this
several times with different line-segments.

6.
cise
Devise and
5,
in
the
perform exercises,
measurement
of
similar
to
angles, using
Exer
a large
protractor and circular sectors of paper as measuring in
struments.
7. Try measuring short intervals of time to tenths of
a second by means of an ordinary watch. In order to test
the results, let the period measured be the time of swing
pendulum, and measure by the watch intervals

ten, fifteen, twenty and one hundred swings, find
of a simple
of five,
ing the time of a single vibration from each measurement.
Do
in
the results agree
Have you any
greater confidence
one than in another ?

8.
Familiarize yourself thoroughly with the use of as
many
different kinds of verniers as are available.
Before
10
using the vernier in each case, estimate the fraction of a

unit in tenths
9.
by the
Weigh a small
ance, then on a trip
Compare the
results.
est confidence?
10.
Weigh a
eye.
piece of iron
by means
of a Jolly bal
scale, then on an equal arm balance.

In which result have you the great
Why?
small object several times, with the high
est degree of precision attainable,
pointer method should be used.
on a good balance. The

Are the results all equal ?
CHAPTER
II
ON THE OCCURRENCE AND GENERAL

The term
Errors and Residuals.
7.
been used somewhat indefinitely, and
it
error
will
has so far
be necessary,
before going further, to explain its exact meaning, as well

as that of another term closely connected with
We
have seen that
different
same quantity generally give

results evidently
is,
measurements upon the

results.
These
different
it is very un
the degree of
even
to
correct,
to the number of decimal places) attain
cannot
likely that any of
precision (that
it.
them
all
be correct, and
is
able with the instruments
and method used.
The
differ
ence between the result of an observation and the true

value of the quantity measured
observation.
is
called the error of the
In what follows we shall generally denote
by the symbol s, the

made by q, and the errors
upon which
observations
quantities
they are
of the observations
x,
by
the latter being defined, as just stated, as the difference
which
will
tion
too large or too small.
is
ful to
q,
(1)
be positive or negative according as the observa

The student should be care
remember this definition, and to apply it to such illus11
12
trations as the following
and the
result of a
If
a line
is
exactly 437 feet long,
measurement upon
it is
436.2 feet, then
While we cannot ordinarily obtain the
true value of a
the error
is
436.2
437
= -
0.8
ft.
measured quantity from one measurement, nor even by

averaging
many measurements,
the
method
what
is
called the most probable value,
which
approximation to the true value that the

vations
is
capable of yielding.
of least squares
means
furnishes us, in the latter case, with a
of calculating
the closest
is
series of obser
familiar illustration
is
that of a series of direct observations upon a single quan

tity, in
which case the most probable value
is
simply the
arithmetical average of the several results.
Having obtained the most probable value from a series

of observations in the manner hereafter to be explained,
if
we now
subtract
it
a series of differences
from each measured
known
result,
we obtain
as the residuals corresponding
The most probable value

being denoted by m, and any observation by s, the resid
to the respective observations.
ual corresponding to s
is
m.
(2)
Thus, the residual bears the same relation to the error that
the most probable value bears to the true value.
ber of observations be very large,
very precise, then the most probable value

very close to,
though
never equal
in that case the residuals will
responding true errors.
If
the
num
and the observations be
to,
may
be very,
the true value;
and
be equally close to the cor
It is
worthy
13
of note that, since the true value of a
quan
terms of any arbitrarily selected unit is always an

incommensurable number, while the most probable value
tity in
is
commensurable,
is
it
follows that the error of
any observa
incommensurable, while the corresponding residual

commensurable. The true value and the errors are con
tion
is
sequently forever
unknown and
figure only in theoretical
discussions (with such exceptions as have been noted)
and we deal
in practice only with
their close approxi
mations, the most probable value and the residuals.
8.
Classification of Errors.
It is
now very important
to point out that errors of observation
naturally into
two
distinct classes,
may
be divided
whose occurrence, and
the methods of dealing with which, are entirely different.

First
we may
consider those errors which arise from
manner
causes that continue to operate in the same
throughout the series of observations, and which
may
be called persistent or systematic errors. In

cases, persistent errors not only occur in the same
therefore
many
manner, but have the same value, throughout the investi

gation, and they may then be called constant errors.
The
causes of persistent errors, which are often
to the observer and
may
in
many
known
cases be eliminated or
avoided by methods presently to be explained, may be,

for the most part, looked for under one or another of the
following heads.
a.
used
Incorrect Instruments.
may
not be true.
The instruments
For example,
if
or scales
a 100-foot tape
14
is
actually only 99.99 feet in length, every measurement
on a
line
made with
that tape will tend to give a result one
ten-thousandth too long, no matter how many times the

observation is repeated; or if a clock used in scientific
work gains one second a day, every measurement on an

interval of time
made with that clock is just the correspond
ing fraction too long.

positive or negative
b.
(In each of these cases,
is
the error
?)
Imperfect Setting of Scale.
Owing
to carelessness
on a measuring instrument, though

truly graduated, may be displaced from its proper
position by a small amount. This is well illustrated by
or accident, the scale
the mercurial barometer, on which the scale must be
adjusted at each reading, to allow for the
mercury
in the reservoir;
rise
and
fall of
and a clock which, though
running at the proper rate, has been set a little ahead or

behind the true standard time, is an analogous case.
c.
Defective
No
Mechanism.
instrument
is
absolutely
and every instru

ment of precision must be frequently tested if we would
The arms of a balance are
rely upon the results of its use.
perfect from a mechanical standpoint,
never really equal, and, what
is
worse, they are continually
changing their relative length, owing to changes of tem

Nor has it been found possible to construct
perature.
a clock that will run with absolutely constant rate, even
at a constant temperature
and
in a
vacuum.
In very delicate instru

ments, such as the balance or the aneroid barometer,
the indicator frequently comes to rest, on account of
d.
False Indicator Settings.
a false position.
friction, in
barometer,
to
make
may
e.
it
15
In the case of the aneroid
often suffices to tap the dial gently, in order
the indicator assume
its
true position.
The same
be said of the magnetic compass-needle.
Known
It is often the case
External Disturbances.
that persistent errors are introduced by external causes
whose nature
avoided.
is
well understood, but which cannot be
Thus, a heated body under experimental in
vestigation always radiates
elaborate precautions
or tape
f.
is
some heat,
in spite of the
and the length
of a
most
measuring rod
certain to vary with changes of temperature.
Personal Equation and Prejudice.
Every observer
which cause
exhibits peculiarities or habits of observation
him
same
to have a tendency toward persistent error in the

direction.
Thus, one observer
may
continually
overestimate in the estimation of tenths, another will
estimate;
a time observer requires a certain
under
definite
is, to obey a signal

This unconscious, persistent error on the
interval to respond to a stimulus, that

of
any
sort.
part of an observer
is
called his personal equation.
Somewhat analogous
to personal equation
is
what may
be called prejudice. After an observer has made one

of a quantity on a fixed scale, and made the
measurement
estimation of tenths, there

to allow his
first
is
a natural tendency for
him
estimation to affect the subsequent ones.
This difficulty is often met with in the use of the vernier,

where it is necessary to judge as to which line coincides
most nearly with
The second
its
fellow on the scale.
class of errors referred to at the beginning
16
whose causes are temporary,

one
observation, and disappearing
existing through only
Such errors
entirely upon a slight change of conditions.
of this article comprises those
are not recognizable, and sometimes not even suspected,

until their existence
is
demonstrated by the discrepancies
between successive observations when

ances have been eliminated.
all
These are
known
known
disturb
as acci
dental errors.
Accidental errors
Those
a.
may
and
Due
to
may
also be subdivided, as follows.
External Causes.
Accidental errors
from causes entirely foreign to the observer

complex a character as to be incapable of analysis.
result
of so
For example, in sighting a mark with a surveyor s transit,

a sudden gust of wind may imperceptibly sway the in
strument for a moment, or someone may, without the ob
server s knowledge, knock against the tripod and jar the
telescope slightly out of place.
In making delicate
place unexpectedly in the earth

affect
momentarily
mag
such rapid changes as often take
netic measurements,
magnetic
field
the equilibrium of the needle.
may
In
sighting at a star with a telescope, currents of air in the
upper atmosphere may cause it to waver and appear for

a moment to one side of its mean apparent position. In
using a balance, the zero of equilibrium
may
change
slightly during the course of a single weighing, owing,
perhaps,
to
an
unsuspected
fluctuation
of
tempera
all
kinds are
which are
of greater
ture.
It will thus
affected
be seen that observations of
by multitudes
of such causes,
or less importance, but which
all
17
tend to affect the accuracy
of the results.
b.
Aside from per

is sub
Accidental Errors of Judgment.
and prejudice, the observer himself
sonal equation
ject to fluctuations of
of his instrument
judgment, both as to the adjustment
and as to the estimation
attempt to analyze in detail the causes
of tenths.
An
of these internal
tendencies to err in judgment would belong to the realm

of psychology
them the
but we
may mention
as prominent
among
influences of imperfect vision, optical illusion,
inattention and fatigue, the last mentioned cause probably

affecting the others in a very large degree.
Some
of the
with persistent
methods commonly employed in dealing

errors are briefly mentioned in Art. 10.
however, the study of accidental errors, and of the

laws which are found to govern their occurrence, that
It
is,
constitutes the special office of the
9.
method
Entirely distinct from
Mistakes.
of least squares.
errors,
in
the
sense heretofore used, are those inaccuracies which are
due purely to carelessness, and which should properly be
They consist in such blunders as reading

number on the scale, reading one number and
called mistakes.
the wrong
putting another
ward instead
ing.
is
in the notes, reading a vernier
etc.
When measurements
checking
is
back
making a miscount in timing
Mistakes are usually easily detected,

no remedy except vigilance and careful check
a pendulum,
and there
down
of forward,
are
a simple matter.
made more than once
the
18
General Methods of Eliminating Persistent Errors.

In Art. 8 are enumerated several causes of persistent
10.
errors,
with illustrations of each.
their discussion
Though
does not properly belong to the general theory of errors,

it
may
not be out of place to describe here some of the
methods commonly employed
in
especially as the theory of errors
frequently applied in
the processes of correction here referred
ment
with them,
dealing
is
The
to.
treat
of the several sources of persistent errors will be
taken up in the same order as they are mentioned in

Art. 8,
and designated by the same
letters.
Adjustment and Standardiza

never certain that an instrument measures
Incorrect Instruments.
a.
As
tion.
it is
in true units,
the results of
made
it is
necessary to test
before relying
upon
may in some cases be

measurements.) An instrument may
its use.
long after the
it
(The
tests
sometimes be adjusted correctly, and remain so; more

commonly it gets out of adjustment again, from wear or
other causes.
or impossible.
tion,
Actual adjustment may often be inconvenient
more approved
practice
which will apply to nearly every
case.
is
standardiza
This consists in
comparing the instrument with a standard and determining

the true value of each of
its scale divisions
or units,
and
then, instead of trying to adjust the instrument, simply
making the necessary corrections on the observations.

(Where standardization extends over a whole
is
commonly
called calibration.)
seldom corrects his clock
scale, it
Thus, the astronomer
he simply determines
from the stars at intervals, and thus deduces
its
its
error
error in
which
rate,
is
all
19
the information needed at any time.
Laboratory weights are seldom correct when purchased,

and moreover they lose or gain weight by wear or corro
hence they should be compared from time to time
sion
;
with standards kept for the purpose. Numerous

tions of the kind will occur to the reader.
b.
The
Differential Method.
Imperfect Setting of Scale.
due to imperfect setting
error
illustra
of the scale
may
often
be eliminated by the differential method, which consists

in reading the position of the indicator
at zero, then again
when
it is
when
it
should be
by the quantity to
This method
affected
be measured, and taking the difference.

applies only
the scale.
when the scale divisions are equal throughout

The process is one very generally employed,
has further advantages than the one here stated;

very frequently it is the only method practicable. The use
as
it
of a level
and
leveling rod in surveying illustrates the latter
point, as does almost
any kind
of
one wishes to weigh a portion of

subtract the weight of the
empty
and contained
liquid.
of the vessel
Defective
Mechanism.
mental errors
often be
c.
may
comparator and when

liquid, he must needs
;
vessel
Compensation.
made
and automatically disappear.

is by far the best method
done,
it
which the
is
the process of
Instru
to react against
selves
simple example
from the weight
"
When
can be
of elimination.
double
effect of inequality in the
this
them
arms
weighing,"
A
in
of the balance
removed by weighing with the object first on one pan,

then on the other, and taking the mean. (Strictly, the
is
20
mean
geometrical
should be used.)
resting
upon an imperfectly adjusted
well as
if
a spirit
If
level,
base, be
simply
to
end
the
between
the
two
reversed,
end,
half-way point
positions of the bubble will indicate its true position as
were in adjustment.
it
The graduated
circles
used on surveying instruments, spectrometers, and the

like, are usually provided with two diametrically opposite
from the vernier system
verniers, so that the error arising
being out of center with the circle
on taking the mean
may
itself
of the readings of the
two
disappear
verniers.
In using a galvanometer it is well to reverse the current

and read the deflection both ways on the scale. An in
teresting application of the
unknown
Rumford
method to the elimination
external disturbance
is
the scheme devised
of
by
for neutralizing the effect of radiation in calori-
metric measurements.
preliminary experiment
by what amount the temperature
to determine
rimeter will be raised

so adjusted that
and then the
it is
initial
is
made
of the calo
temperature
is
about the same amount below the
temperature of the surrounding air at the beginning of

the experiment as
cally the same
it is
amount
above
half of the operation as

d.
False
Indicator
at the close, so that practi
it
of heat is
is
absorbed during the
radiated during the last half.
Settings.
where the indicator comes to
Oscillation.
In
may
often be removed
allowing the indicator to
come
to rest at
while
still
advantage
oscillating.
of saving
cases
rest in a false position,
to friction, the difficulty
it
first
all,
due
by not
but reading
This method has the further
time in such instruments as the bal-
21
ance and undamped galvanometers or magnetometers.

In order to compensate for diminishing amplitude, one
more reading should be taken
at one extreme of the swing
than at the other, as in the following balance pointer

readings and reduction:
RIGHT
LEFT
13.1
7.8
8.0
13.0
8.1
2)26.1
13.05
2)21.02
10.51
This result
is
True reading.
much more
quickly obtained and more
accurate than one obtained by letting the pointer come to

rest.
e.
Theoretical Corrections for
When
ances.
the
manner
in
Known
External Disturb
which external disturbances
known, and their magnitude determined, the

operate
errors due to them are eliminated by simply applying the
is
The temperature and
corrections.
computed
stretch
corrections applied to the steel tape in precise chaining,
and the temperature corrections necessary with instru

ments, such as the barometer and pyknometer, depending
upon the density

vessel,
are
Rumford
amount
of a liquid or the capacity of
familiar
examples.
in
of
employing
using the calorimeter, the
compensation
minute
of radiation per
Instead
a hollow
may
be previously noted
22
and allowed
The
for in the reduction of the results.
re
fraction error in the observed altitude of a star, or in long
range leveling, the vacuum correction in weighing, etc.,

are further familiar examples.
It is for the purpose of
obtaining data for such corrections that
many
investiga
phenomena under varying

indeed, this work constitutes a
tions of the behavior of physical
conditions are carried on
large part of quantitative scientific research.
Corrections for Personal Equation
f.
Personal equation
mining by means
may
and Prejudice.
be eliminated, either by deter
of specially devised experiments
the personal equation of the observer
is
what
for a given kind of
measurement, or by arranging matters so as to make the

personal error act in opposite directions in the two halves
of the observation
that of employing a
or by a very different method,

number of different observers on the
same measurement, whose
errors will tend to
compensate
in the long run, like accidental errors.
The
effect of prejudice
may
often be avoided
by
altering
Thus, when repeatedly using the differen

method, the whole measurement may be shifted each
the conditions.
tial
time to a different part of the
scale.
The oscillation method
not subject to prejudice, since, though the true reading

may be the same in the successive observations, the oscilla
is
tions approaching
will
it
will
not be.
An
experienced observer
not allow prejudice to influence him to any great extent.
11.
Exercises Leading to an Understanding of Error

Before attempting any introduction to the
Distribution.
methods
it is
of dealing
23
with accidental errors in measurement,
necessary that the student recognize the existence of
a law governing their occurrence, and become to some

extent familiar, through experience, with the operations
To
of that law.
this end,
introduce at this point a
it
is
deemed worth while to
number
or experiments, in which the
of laboratory exercises
phenomenon
to be studied
is
the distribution of errors as governed by the law of chance.
The term
exercises
any
"
"
laboratory
refers to the
be performed at one
may
method only; the
study table without
special apparatus.
No
better analogy to the behavior of accidental

can
be found than in the manner in which shots
errors
1.
fired at
a target are found to distribute themselves with
respect to a point fired at.
To
illustrate
this experi
mentally, take a sheet of ordinary foolscap or other ruled
paper and with a black pencil make the ruled

the middle of the
sheet
heavier
line nearest
than
the others, so as to
be distinctly visible
a few feet away.
Lay the paper on a
board
or
smooth
book, and place
it,
face upward, on the

floor.
Take a rather long
FIG. 1
pencil lightly
between the ex
tended finger-tips of both hands, and standing with the

eye directly over the black line on the paper, hold the
24

downward, over the
pencil, point
drop
it
line,
and endeavor to
so as to strike the line with the descending pencil-
In other words,
point.
make
the central line a target;
the shots will be self-recorded by the dots on the paper.

at least a hundred shots in this manner, each time
Take
trying with
all
possible skill to hit the central line.
Having
done
this, prepare another sheet of paper ruled off in a

similar manner (ordinary coordinate paper will do) and
plot on it a curve whose ordinates represent the relative
number
ment
found to have struck in each compart

and whose abscissas represent the
of shots
of the ruled target
distances of the respective compartments from the central
In case a shot appears to have struck exactly upon
line.
one of the
lines, assign it
to the compartment on the side
toward the center.
Can you think
of
any influence that might,
in this ex
periment, be analogous to a persistent error in measure
ment?
What
effect
would
it
have on the curve?
Keep
the data for future use.

2.
On
a sheet of smooth paper, draw a line with a hard,
sharp pointed pencil and

foot apart.
The
exercise
mark two
is
points on
it
about a
to measure this line with a
.metric scale to hundredths of a centimeter, estimating the
hundredths as tenths of a millimeter.

prejudice,
it will
In order to avoid
be well to place a third point somewhere
between the others, and measure the line in two segments,

a and 6. Now measure a and b alternately, using the
differential method, until each has been measured, say, a
hundred times. Add the corresponding pairs of values
25
and record the sums as the measured lengths of the line.

Find the mean of the hundred values to the nearest hun
dredth of a centimeter, and record the departure from it
of each of the observations, plus or minus.
These de
partures are the residuals of the observations (Art. 7).
It
be noticed that a large number of residuals have the
will
Determine how many there are of each

value, separating positive from negative, and plot a curve
whose abscissas represent the values of the residuals and
same
value.
whose ordinates represent the numbers

those respective values.
used
0.1
for example,
mm.
of residual,
on the
of residuals having
convenient scale should be
abscissas, let 1
and on the
ordinates, let
be represented by a millimeter.
Keep
cm. represent
each residual
the data for future
use.
What change would have
to be
made
in the curve
if
the abscissas and ordinates were the values and numbers,

respectively, of the true errors instead of the residuals,
supposing that there

3.
The preceding
is
any means
exercise
may
of
knowing the former ?
be varied by using, for
the measured quantity, an angle of exactly 180, measuring
two segments with a protractor to tenths of a degree.

In this case the true value, and hence the true errors, are
it
in
known.
4.
Do
Keep
the data.
the curves obtained from the preceding exercises
bear any resemblance to each other ?
Construct a smooth
curve which seems to be typical of them.

resemble any familiar geometrical form?
2"^,
taking 10 cm. as the unit
for
Does
this curve
Plot the curve
both abscissas
26
and ordinates and assigning to x the successive values

0.1, 0.2, 0.3, etc., both positive and negative.
6.
From
0,
the results of the foregoing exercises, does
there appear to be any relation connecting the magnitude

of
an error with the frequency
errors appear to occur
Can
of its occurrence?
Do
you assign any reason for such a relation?

any more frequently,
positive
in the long
run, than negative errors, or vice versa ?
12.
Remarks on the
The curve
Distribution of Errors.
to which the preceding exercises have introduced us
commonly
called the probability curve,
is
though a better
name would be
curve
of
the
departures,
as will appear later.

Superficially
it
some
what resembles the

"
witch,"
case being
Fig. 2.
typical
shown
in
The student
must not expect that
G<
any curve plotted from the results of such experiments

as the foregoing will be smooth and regular, like the curve
here shown;
actual curves are broken
But the greater the number
and
irregular.
of observations or data, the
nearer will the actual departure curve assume the smooth,
symmetrical form assigned by theory.

The results of experiments, as we have seen, and theoreti
cal considerations, as will appear,
both point to the follow-
27
ing facts regarding the distribution of accidental errors,

all of which may be deduced from an examination of the
curve.
1
The frequency with which an accidental error of given
magnitude occurs depends upon the magnitude of the error.

2. Large errors occur less frequently than small ones.
The error distribution
3.
and
negative errors of the
symmetrical; that
is
is,
positive
same magnitude occur with
the
same frequency.
in
Though
these laws do not of course apply absolutely
any one
case, yet
they express the general tendency of
error, and, in fact, the general
tendency of
all
accidental
departures from the normal or mean, as, for example,

the statures of individual people as compared with the
average stature of the race.

the
is
number
of observations
regarded as
infinite,
In theoretical discussions,
made, or
of data considered,
and the curve as
strictly
sym
metrical.
In the case of measurements, with which we are here

concerned,
if
the results are affected by persistent error
from any source, they
will
be found to cluster about the
theoretical most probable value of the
measured quantity
instead of the true value, there being
now an
difference
between the two.
now becomes a curve

a
little
error
is
appreciable
The whole curve
of residuals,
and
is
of errors
merely shifted
to one side or the other according as the persistent

positive or negative.
If,
exercise of Art. 11, the scale used
for example, in the second
had
its
spaces slightly too
long, the whole curve would be shifted a
little
in the
28
negative direction, simply because each observation tends

to undervalue the line measured on account of the defect
in the scale.
From
this consideration
it is
clear that
when, as
is
really
always the case, the true value of the measured quantity

is not
given by
the measurements,
of
study
the
curve of residuals
will reveal
nothing
as to the presence
or absence of per-
The
sistent errors.
law of probability
error is con
FIG. 3
of
cerned only with accidental errors, that

causes
are of temporary duration,
is,
those whose
the result, as
we
say, of pure chance.
13.
On
Precision.
comparison of the results of
differ
ent sets of measurements, even upon the same quantity,

it is found that the error curve is not of constant form.
Every gradation
is
met with
to high, pointed ones.
when we make
(Fig. 4),
several series of
same quantity by
from low,
This peculiarity
may
flat
curves
be observed
different methods.
The
variation
is
easily interpreted.
Compare,
for example, curves
there are nearly as
many
and D.
In the case of
large errors as small ones.
For shots
fired at a target, this
manship or long range

judgment,
crude
render the work
in
would indicate poor marks

it means random
measurement,
instruments,
difficult.
29
or
circumstances which
In the case of D, on the other
hand, the number of large errors
very small, the great

body of results being crowded closely about the mean
and
with
siderable
con
definite-
From
ness.
is
its
indicating
position
is
this it
clear that the
form
of the error or resid
ual
curve
depends
upon the precision

with which the ob
servations have been
made.
To
illustrate
what
meant by preci
sion, let two parties
is
of observers each
FIG. 4
make
a set of measurements on the dis
tance between two stakes, the one with a ten-foot pole, the
other with a steel tape.
from one
set
may
The most probable value deduced

much from that deduced
not differ
from the other, but the residual curves plotted from the
two sets of results will show considerable difference of
precision,
mainly on account of the larger number of
times that the ten-foot pole must be laid
consequent greater
liability to error.
down and
its
30
The form
of the residual curve
may
therefore be used as
a test of the efficiency of an observer, an instrument or

a method of measurement. It will be seen later that the
same
can be applied by means of mathematical
test
formulas, without the labor of plotting the curve (Chapter
VIII).
14.
Mathematical Expression of the Law of Error.

existence of some law governing the distribu
The evident
tion of errors leads us to inquire
whether
tion.
it is
Some
what that law
is,
and
expressible by a simple mathematical rela
of the facts concerning the behavior of errors
have already been deduced but the theoretical expres

sion of the law itself, and even the very language in which
it is expressed, must be reserved until the student has
;
reviewed
some
of
theory of probabilities
the
fundamental principles of the
and has been introduced to some
of
the special problems in probability upon which the theory
found to depend. The following chapter

devoted
to this subject.
therefore,
of errors is
is,
CHAPTER
III
ON PROBABILITIES
15.
Fundamental
that one thing
is
It is a
Principle.
more
likely to
common remark
happen than another.
In speaking thus, one concedes that either of the two

events may happen, and attempts no prediction as to
which will happen,
if
either
yet he recognizes a preponder
ance of the likelihood of one event over that of the other.

In the kind of magnitude here recognized, that is,
likelihood or probability, there is, in the great majority
of cases, no means of measuring or giving numerical ex
pression to
its
relative degrees.
It
is
said
that corn
growing on low ground is more likely to be caught by frost

than that on high ground, but there is no means of telling
how many times more

It
is
possible,
likely
it is.
however, to give such a definite meaning
to the term probability that the relative probabilities of
some simpler events may be calculated and expressed.

framing such a definition,
it is
In
necessary to recognize an
important principle in the operation of chance, governing

the behavior of events whose causes are at least partly
manifest, and lying at the foundation of the whole course
of reasoning that gives rise to the idea of
probability.
31
mathematical
32
The
is this.
principle
// a number of different events
are equally possible as regards
constant
conditions
(that
no persistent reason why one should occur

is, if
rather than another), and all are repeatedly given oppor
tunity to occur, they will in the long run occur with
there is
equal average frequency.
pressed by saying that
if
The same principle may be ex

we observe events occurring with
we conclude
equal frequency,
that the constant conditions
under which they occur are uniform.
The
If
principle
a die
is
is
by the throwing
well illustrated
of
cubical,
exactly
of dice.
homogeneous material
and the spots do not shift the center of

gravity to one side, and if it be cast a great number of
times absolutely at random, each face will come up, on
"
loaded
(not
")
the average, one throw out of
(Of course these ideal
six.
conditions are not realized in practice.)
We
are so accustomed to the operation of this law of
probability in
daily experience
matter of course,
existence
is
chance,
16.
it
is
taken as a
like the force of gravitation;
really a mystery.
admit that there
that
is
We
yet
its
are here obliged to
a law controlling the operations of
the one thing that would seem to obey no law.
Definition of Mathematical Probability.
numerical significance
may now
Definite
be given to the probability
of occurrence of certain classes of events.

If
an event
may
occur in a equally possible ways,
and at the same time

are presented in
all
equally possible alternatives
(including the a
ways
in
which the
ON PROBABILITIES
event
happen), then the probability of the event in
may
question is defined
as the ratio
P=
That
is,
33
(3)
1|
l-
there are a chances favoring the event out of a
total of b possible chances

set forth above,
if
a great
and according to the
number
of trials are
principle
made, the
event does happen, on the average, a times out of 6.

As an example, let us express the probability of draw
draw
ing an ace from a deck of fifty-two playing cards, the
ing being done absolutely at random. Any one of the

fifty-two cards may be drawn, so that the total num
ber of alternatives
is
be drawn in only four ways,

of
An
fifty-two.
viz.,
ace may, however,
by drawing the ace
spades, the ace of clubs, the ace of
ace of diamonds.
Here, then, b
52, a
hearts
4,
the
or
and the
drawing an ace is ^, or T^.

What would be the probability of drawing a red ace?
probability of
Of drawing the ace of diamonds ?

All problems in probability
may
be solved by the appli
cation of the definition expressed in equation
such direct application would be very

complicated
cases,
and
special
therefore to be devised which,
difficult in
rules
But
(3).
the more
and formulas are
when properly
classified
and applied, greatly simplify such problems.
From
the definition,
it
follows that probability
is
purely numerical ratio, and depends upon no unit of
measure.
The
Moreover,
this
ratio
cannot
exceed
unity.
probability unity would denote certainty, since
if
an
34
event
may happen
in
n ways, and only n
possible, the event must happen.
that
the probability of an event
if
of its failure to
happen
For,
to
is
alternatives are
this it follows
p, the probability
is
P =
if
From
P-
(4)
the event can happen in a ways out of
happen
a ways out of
in b
6,
6, it
can
fail
the probability of failure
therefore being
*
.
More
generally, the
alternatives
The
is
sum
of the probabilities of all possible
unity.
probability zero, on the other hand, implies im
It may be interpreted as meaning that there

or in cases
no way for the event to happen, i.e., a =
where the total number of alternatives is infinite, or at
possibility.
is
least extremely large, while the event in question
may
very few ways, the zero or infinitely

small probability denotes impossibility or at most only
extremely remote possibility. But the distinction be
happen
in only a
tween absolute impossibility and the case in which the

possibility is only remote is of some importance, as will
be seen, in the theoretical discussion of the distribution of
errors.
17.
Permutations.
The
solution of problems in prob
ability involves the determination of the
in
which an event can occur, as well as the
possible alternatives.
number of ways
total number of
In very simple cases this
may
be
ON PROBABILITIES
done by inspection.
For example,
if
35
one
is
expecting the
arrival of three different persons, A, B, C,
it is
easy to
determine the probability of their coming in the order
There
named.
which they
is
The
orders
obviously six different
may come;
CAB, CBA.
ABC
are
namely,
ABC, ACB, BAG, BCA,
probability of their coming in the order
therefore J.
But
instead of three, and the
let
there be a hundred persons
number
of orders
becomes so
enormous as to be unmanageable by inspection.

must then resort to the use of general formulas.
The linear permutations of a number of

ferent
ways
which the things
in
or in which they
in
may
may
We
things are the dif
be arranged in a row,
There
occur in order of time.
are,
for example, six linear permutations of the letters A, B, C.
There
tions of
is
a general expression for the
number of permuta
different things, derived easily
by the following
Of one thing, there is evidently but one per

Of two things, since either may come first,
reasoning.
mutation.
two permutations. Of three things, any one

may come first, and with a given one coming first, there
are two arrangements of the two remaining; therefore
there are
the
number
of
permutations of three things
is
6.
For four things, by the same reasoning, the number is

4 X 3 X 2 = 24. And in general, the number of per
mutations of
things
is
PQ = Q(Q-l)(Q-2)...3-2.1=Q!
We
have here assumed that none of the
duplicates.
(5)
things are
Let us now take a case where there are
36
AABBBCCCC.
duplicates, as in the group of letters
we
distinguish between the different
same as
of course, the
But
if
we
if
etc.,
s,
the letters were
consider one A, for example, the
the case
all
If
is,
different.
same as another,
and permute without regard to which of them is being

used in a particular place, the number of permutations
is
It will
less.
exercise that
kind,
if
be easy for the student to show as an
there are, in a
number
of another, r of another,
number being Q =
m+
+r+s+
Thus
A s,
for the
three
If there is
m,
n,
above
- --
distinguishable permutations of the
pQ (m - =
of things,
the
things
number
of
is
2J
(g)
mlnlrlsl"
set of nine letters, of
and four
m of one
of another, etc., the total
the
s,
number
which two are
is
only one thing of a kind in the group, so that

are each unity, (6) becomes equivalent to (5).
The different groups which can

18. Combinations.
be formed from a number of things, taken so many at a
time, are called combinations.
of the three letters A, B,
AC, BC.
If we further take
The
different combinations
taken two at a time are AB,
into account the possible
tions of each combination,
permuta
we have what may be
called
the permuted combinations of the series of things considered.
ON PROBABILITIES
37
Thus the permuted combinations of A, B, C are AB, BA,

AC, CA, BC, CB. It is easier to derive first the general
formula for the number of permuted combinations.
Let the number of permuted combinations of Q things
If
taken n at a time be designated by the symbol P(7g (n)
of
the
one
a
two
at
taken
are
time, any
Q things may
they
.
be taken as the
things
may
first,
and any one
of the
remaining
be taken as the second, so that
taken by threes, any one of the Q (Q

1) permuted
combinations of two each may constitute the first two,
If
by any one
the third. Then
of the
followed
By
2 remaining things as
continuing the same reasoning until there are n things
taken at a time, we readily deduce
=Q(Q -
PCQ
If
- 2)
.- to
n = Q, this becomes identical with
are
express
taken n
now
the
number
tions include not merely those

all
things.
(5),
factors.
(5), since all
(7)
the things
of combinations of
Q things
at a time, without regard to their arrangement,
necessary only to note that the
but
permuted at once.
To
is
l)(Q
(n)
it
permuted combina
PCg
made up
of different things,
the permutations of each of the groups of n different

Since
n things
there are only
are
as
n\
permuted
many
in
different
ways
combinations as permuted
38

That
combinations.
is,
M _Q(Q -I)
r
CQ
"-ton
TT
As an
factors
illustration of this problem, let us find
Qv
how many
hands at whist, each made up of thirteen cards,

could be drawn from a pack of fifty-two cards. Here
Q = 52, n = 13, and the solution is
different
Cfi2(13)
= 52-
Then the
by
51
L
-SO--- 40 =
6 35 j013) 559 j6 oo.
o
lo
probability of drawing
any one
hand
specified
is,
the exceedingly small reciprocal of this
definition,
number.
As a
final
problem
of things, the
in combinations, let there be s series
number
being Qi, Q%,
of things in the respective series
to determine
how many
different
combinations can be formed by taking one thing from

each series.
The number
of combinations of
be formed from the

the Qi things in the
first
two
two each which can thus
series is QiQz, since
first series
each of
can be successively com
bined with each of the $2 things in the second. Bringing

in now the third series, each of the QiQ 2 combinations just
considered
may
be combined with each of the
Q members
s
making QiQzQs combinations; and so

on.
Clearly, then, the number of combinations that can
be so formed from the s series is the product
of the third series,
PROBABILITIES
OAT
For example,
let
39
there be three series of letters
BZ
C%
A%
A,
Bs
A,
The number
form
of combinations of the
be selected from them
is
ABC
that can
Let the stu
24.
dent write these combinations.
Two
Probability of Either of
19.
the probability of an event

Pb, that of an event
is
More
p a that
is
then
etc.,
c,
or
of
Events.
an event
it is
If
is
easy to show
that the probability that one or another of these events will
happen
is
pa
+ Pb + p +
c
being understood that
it
only one of these events can happen.
event
For, suppose the
A may happen in a ways, the event B in b ways, etc.,
and that the
number
total
of
alternatives
T.
is
T will
(In
be greater than the sum of a, 6, etc. that is,

general,
it is not necessary that any one of the events A, B, etc.
shall
happen.)
Then by
definition, the probabilities of the
respective events are
Pa
If
we
A, B,
designate
etc.
since the
by
etc.
j,,
?fc=y,
X the event of
some one
of the events
happening, without specifying which, then,

6
of ways in which
can occur is a
number
the probability of
is
do)
40
As an example,
let
The
^%
of copper,
probability that a metal ball will be
^-
-f
=
2*-
-f-
two
wood, seven of lead, six of rubber, one of

and let one be drawn at random.
of glass, five of
ivory and four
there be in a bag three balls of iron,
be an iron
since a metal ball
is
!>
drawn
drawn
then
is
if
the re
a lead ball or a copper ball.

This principle of additive probabilities for alternative
events is made use of in estimating premiums on so-called
sult
"
"
life
joint
ball,
insurance policies.
Quite a different problem
Events.
probability that
As
shall occur.
A, B, C,
pa
by
Concurrence of Independent
Probability of the
20.
is
that of finding the
a specified set of independent events
before, designate the respective events
etc., their
Pb, etc.
Z.
all of
respective separate probabilities
and designate the event
Suppose the event
A may
B
a ways out of a alternatives,
by
by
of their all occurring
occur independently in
in b
ways out
of
alter
natives, etc., so that
Pa
It is of course
etc.,
= a~,
a
b
= -,
etc.
/3
understood that when
all
are given opportunity to happen,
alternatives connected with

/3
&
the events A, B, C,
some one
A will happen,
alternatives connected with
will
of the
some one
happen,
of the
etc.,
but
The total number of

number of combinations
that only one of each can happen.

possible
outcomes
is
that can be formed
therefore the
by
selecting one
from each group of
alternatives, namely, the product a(3y
(9).
Likewise,
ON PROBABILITIES
the
number
etc.,
can
all
of different
occur
the probability of
of the event Z,
P*
the product abc
is
all
It follows that
occurring, that
a
-
is,
the probability
c
b
--"=
PaPbpc
ajSy
That
which the events A, B,
is
abc
in
ways
ft
to say, the probability of the concurrence of two or
is
more independent events
is the
product of the probabilities
of the respective events considered separately.

is
of course less
To make
known
than any one of
that a person
will
visit
finding
is
and 6
known
the place at
of finding
of finding
spend
five
it is
hours in a certain
+ T2 = f
3
when these hours will be. If

any random moment, the probability
there at that
them
P.M.,
as to
A there at that
moment
moment
is
both there at that
But the probability

5
2
will
and that another person

spend three hours there during the same interval,
but nothing
we
This product
its factors.
the meaning of this clear, suppose that
place between 6 A.M.
,.,.,>.
(11)
is -f^
Y2~-
>
the probability of
Then
moment
of finding either
the probability
is -f%
or
X
B
&
-f$.
there
is
Let the student analyze this problem more
showing how the values stated for the probabilities

can be deduced from the definition of probability.
closely,
21. The Coin Problem.

Suppose that the result of
an experiment may be either one of two things, A and B
which are equally likely to occur, and that the result must
?
be one or the other, but cannot be both.

of either result is then
The
probability
Let us determine what
is
the
42
coin
example,
if it
ability that
The
be tossed a hundred times, what is the prob

turn up heads thirty-eight times and
it is
= 100,
Here Q
required probability
more,
times, that
it will
sixty-two times?
tails
n times one way and Q

n times the other.
tossed at random illustrates the problem; for
result
it will
the experiment be performed
if
probability,
is
ft
a function of n
evidently the same function of
= 38
;
(or
62).
and, further
n that
it is
of n.
The
result
first
thing to determine
A may happen
the coin, heads
may come up
a large variety of ways
and 60
T., in order,
would
in
is,
n times out
how many ways
38 times and
for example,
fulfill
the
In 100 throws of
of Q.
tails
62 times in
H., 2 T., 37 H.
the condition
or,
equally
The number of ways

in which A may happen n times and B, Q
n times is
seen
to
be
to
the
number
of
readily
equal
distinguishable
n
permutations of Q things, n being of one kind and Q
well, 8 H., 5 T., 30 H., and 57 T.
which
of the other (Art. 17),
(n,
Q-n)
is
gj
n
Or,
it is
equal to the
number
- n)
(Q
12 )
of combinations of
things
taken n at a time, since out of the totality of Q events,

the n events A may be selected wherever desired. Hence
another expression for the required number is equation
(8), which the student may readily show to be equivalent
to (12).
We
shall use
equation (12).
Next we must determine the
total
number
of possible
ON PROBABILITIES
This
alternatives.
may
be done by adding together the
values of the expression (12) obtained

integral values
from
43
by giving n
all
These are tabulated below.
to Q.
pQ (n,Q-n)
9(9-D
2!
9(9-D(9-2)
TT
9(9-1)
0_ 2
2!
The
to
9-1
expressions obtained for
Q- n ) as n varies from
PQ n
<
are at once seen to be the successive coefficients of
the expansion of a binomial with exponent Q, an d their

sum is therefore equal to 2 Q That is,
.
We now
number
of
can happen n times and event B,
by
(12),
(13).
is
and the
The
therefore
2*.
(is)
have the two elements of the solution of the
coin problem, namely, the
total
number
ways
in
which event
times, given
of alternatives,
given
by
required probability of the specified outcome
44
Thus the
probability of the result 38 heads and 62 tails
is
100!
2
The
38! 62! 2 100
reason for introducing the coin problem will appear
later.
22.
Important Exercise.
It is
now very
desirable,
for the
purpose in hand, that the student faithfully per

form the following exercise. Suppose a coin tossed ten
times.
Find the probability of each of the following
possible results
10
10
10 heads and
tails
4 heads and
9 heads and
1 tails
3 heads and
7 tails
tails
2 heads and
7 heads and 3
tails
6 heads and 4
tails
5 heads and 5
tails
8 heads and 2
heads and
6 tails
tails
9 tails
heads and 10
tails
Considerations of symmetry will shorten the work.
Now
plot a series of points, of which the abscissas shall represent
the quantity n
5 (n being the assumed number of heads)
and the ordinates the computed
probabilities of the re
spective results, using a convenient scale for each.
Does
the resulting curve resemble any other curve that has
come to your
Test the theory by actual

or
a flat bone disk, record
with
a
coin,
better,
experiment
ten
throws.
This exercise, if
of
every
ing the outcome
hitherto
carefully performed
notice
and
studied, will assist the student
ON PROBABILITIES
much
to a
45
better understanding of the behavior of error
distribution than he could attain without
it.
As has been
Empirical or Statistical Probability.
23.
before noted, in the majority of the events of
life,
the con
admit of any such analy
ditions are far too complicated to
as has been applied to the problems concerning cards,
sis
balls, coins, etc.
But
it
may happen
that,
when
the con
ditions are sufficiently constant throughout a long series

of observations, the probability of such a
complex event
be deduced from the observed results.
may
this principle that reliance
is
a very important example,
It is
placed upon statistics.
upon
As
we cannot compute, by any
theoretical formula, the probability that a person ten years
old will live to be sixty.
But
the statistics show that
if
out of every 100,000 persons ten years of age, 58,000

live to
be
ability
is
sixty,
0.58.
we may conclude
do
that the required prob
In a similar manner
it
has been deter
mined that the probability that a person sixty years old

will live one more year is 0.97, since 97 per cent, of those
attaining the age of sixty do live another year.
portance of such knowledge,

cal
and
its
The im
bearing on the practi
problems of the world, such as
life
insurance, are
self-evident.
EXERCISES
24.
1.
What
two throws
dice?
of
is
the probability of throwing a six in
a single die?
In a single cast of two
46
How many
~2.
letters in the
arrangements
3.
possible arrangements are here of the
word
ribbons,
their
She has planned to have the
partners
all.
How many
4.
guests
may
she invite
more couples by giving
she be able to distinguish
Four colors?
each guest three colors?

-
same number
by the matching of colored

each guest wearing two colors, there being but five
find
different colors in
Would
distinguishable
word minimum?
hostess wishes to have as guests the
of ladies as gentlemen.
guests
How many
travel?
of the letters in the
How many different football
elevens could be formed
from a squad of fifteen players?

What chance would
one
of
on
have
a
any
player
getting
picked eleven if it
were chosen by
-
6.
In a certain organization there are two candidates
an
for
lot ?
office
and
That
that there will be a tie?
ceive a majority of exactly f

6.
By
What
thirty voters.
is
the probability
either candidate will re
measuring a number of ordinates of the curve
obtained from the target experiment (Art. 11), determine

empirically the relative probabilities of the respective
errors in aim.
7.
Find the number of combinations of three things
eight;
8.
the
number
of
student council
from each
ways may
permuted combinations.
is
to be
of four college classes,
ships are 150, 105, 75
in
and
the council be
56.
made up of five members

whose respective member
In
made up ?
how many
different
ON PROBABILITIES
**
9.
47
Three things are selected at random from eight,

and then another random selection of
then returned
three
is
similarly
made.
What
is
the probability that
the two selections will be exactly reverse permutations

of the
same three things?
A new
10.
janitor has a
bunch
of twenty-eight nearly
What
similar keys, one for each door of the building.

is
the probability of his being able to unlock the
three doors with only one
trial
each?
Solve
first
also
on
the supposition that he marks the keys as he discovers
them.
What
11.
is
four figures, selected at random, will

alike
and the other two
12.
The
number of
have two figures
the probability that a whole
figures alike
following data are taken from the American
Experience Mortality Tables used by
life
insurance
panies in computing risks.
Out
of 100,000 persons ten years of age,
100,000 live to be at least 10




com
48

Plot a curve representing these data.
-
Find your own chance of living to be at
13.
least seventy
years old, using the curve in Ex. 12.
Three men are respectively
14.
Find the probability that they
Two
16.
and
if
30, 27
and 22 years
will all live to
brothers are respectively 25 and 35 years old,
their father
is
60.
The
elder
is
to inherit the estate
living at the father s death, otherwise the
inherit
it
old.
be 60 or over.
and at the death
will, if living, inherit
younger
will
of the elder son, the
the estate from him.
own
own it
younger
Find the prob
ability that the elder son will
the estate five years
hence
ten years hence.
that the younger will
CHAPTER
IV
THE ERROR EQUATION AND THE PRINCIPLE OF

LEAST SQUARES
25.
Analogy
of Error Distribution to
Coin Problem.
measurement
It
was pointed out
is
the resultant of innumerable small disturbances of
in Art. 5 that
an error
in
different kinds, the presence of many of which may not

be even suspected. These disturbances operate, some in
one way, some in the other that is, some tend to produce
;
positive error
and some negative.
depends on the relation
of the
The
number
resultant error
of positive disturb
ances to the number of negative disturbances.

all
are positive, the error will be positive
nearly
all
and
If nearly
large;
if
are negative, a large negative error will result
about the same number are positive as negative,

the error will be small. This does not imply that the dis
while
if
turbances are
all of
tration, suppose
the same magnitude.
we
select
By way of
illus
from a sand-heap, at random,
a thousand grains of sand, and put eight hundred of them

on the left pan of a balance and two hundred on the other.
There
is
hardly a remote possibility that the former will
not very largely overbalance the
hundred on each pan, there

ance one way or the other. And
five
49
latter.
will
be
But
little
if
we put
preponder
this does not imply, by
50
any means, that the

individual particles
Now
there
all equal in weight

some
be ten times heavier than others.
grains are
may
a remarkable and useful analogy between
is
the theory of error distribution and the so-called coin
problem (Art. 21), an analogy that the student has no

doubt already observed. It is easily deduced that the
most probable
result of a
number
of
throws of a coin
that they will be half heads and half

this
normal result
will
degree, so that in one

tain fifty-five heads
is
In general,
tails.
be departed from in greater or less

hundred throws we frequently ob
and
forty-five tails, or less frequently,
This departure from the

sixty heads and forty tails, etc.
normal or most probable result may be looked upon as
a sort of error. Like an error in measurement, it is com
plex in character, depending
dividual throw.
final
upon the
Each throw, head
outcome one way
result of each in
or
tail,
affects the
or the other, just as each small
disturbance, positive or negative, affects the result of an
observation in measurement.
little
consideration of
the two cases will bring out their analogy quite clearly.
We
are therefore justified in assuming that the proba
bility of the occurrence of
magnitude
of the error in
an error
much
is
a function of the
the same manner as the
probability of a departure from the half-and-half result

in tossing the coin
parture.
Hagen
It
is,
is
a function of the extent of the de
in fact,
upon
this line of reasoning that
deduction of the error equation
is
based.
The
deduction is, however, rather cumbersome, and we shall

follow instead the more elegant method due to Gauss.
THE ERROR EQUATION

It
is
may
51
be remarked here that the theory of departures
a very general one and finds application in a large variety
common experience,
of problems of
such as the distribution
of shots
on a target and the distribution
teristics
among
the
members
of given charac
of a biological group.
The Most Probable Value from a Series of Direct

If a series
The Arithmetical Mean.
26.
Measurements.
of
measurements be made upon a
single quantity
under
as nearly constant conditions as possible, the result
is,
in general, a series of different values, each approximating
the true value of the measured quantity.
No
one of them
the true value, however, and it now becomes a matter of

judgment to select, from all possible values, such a one
is
as will
make
the actual distribution of the results appear
An
most natural.
that after
first
all
analogous case would be this
Suppose
the shots had been fired at the target in the
exercise of Art. 11, the central line
aimed at were
erased, and we were required, from the given distribution

we
of the shots, to judge as to where the line had been
;
could do no better than to select a position that, from the
concentration of shots about

respect to
it,
it
and
their
symmetry with
seems to be the most probable one.
Likewise,
measurements, we are aiming at a true

most
the
value,
probable location of which can only be
in a series of
estimated by an examination of the distributed results.
The symmetry of the

the true value
is
distribution of errors in cases
known,
where
as also in the analogous coin
target problems, leads at once to the
and
common axiom
of
52
experience, that the best value to adopt in the case of a

series of direct observations
arithmetical
measured
several
and
mean
their
on a
single quantity is the
or average of the observations.

results
If
the
be designated by Si, sz
sn
the residuals (Art. 7) are re
,
mean be m, then
spectively
si
m,
pz
=
=
s2
m,
pn
sn
Ss
pi
m.
Adding these we obtain

2/o
nm =
(15)
0,
which expresses the fact that the arithmetical mean of

the results
the value with respect
is
metrically placed, the algebraic
being then equal to zero

the
to
sum
which they are sym

of the
and that therefore
differences
this
mean
is
most probable value that can be assumed.
Gauss
Let q
represent the unknown true value of a quantity and let a
series of n measurements be made upon it, the number n
27.
Deduction of the Error Equation.
being supposed very large.
Let the errors arising from

xn
It has
x\, x2
the respective measurements be
been seen that the probability of the occurrence of an

error
is
some
sort of inverse function of its magnitude.
Designating the probabilities of these respective errors

,
by 2/i 2/2,
yn , this fact may be expressed by the
>
equations
THE ERROR EQUATION

=
2/2
2/n
It is the
form of
53
/(*).
this function /(x) that
we
are seeking to
determine.
we do not know the true value q

the observed quantity, and therefore we do not know
Now,
of
as above noted,
the true errors
We may however assume various tenta
x.
tive values for q
and study the
resulting tentative systems

view to selecting that one
which seems most naturally distributed, in accordance
with the notions of error distribution that experience has
of errors, particularly with a
taught
and the
us.
In this sense, therefore, we
may
think of q
errors x as variables subject to our control,
probabilities y will
understanding, then,
then vary accordingly.
we
and the
With
this
are seeking to find that system
of values for the # s which, as a whole, has the greatest
probability.
If
tem
the outcome of a series of measurements be the sys

of errors x\, x%,
as the concurrence of
is
x n this result
,
n independent
be looked upon
events, each of which
the obtaining of one of the errors
may
x.
Then according
by
to Art. 20, the probability of this outcome, designated
F,
is
the product of the probabilities of the separate errors,
namely
Y =
2A2/2
2/n
/fa) -/fe) -/(*).
In order, therefore, that the system of z
s shall
greatest probability, as required, the value
(16)
have the
assumed
for q
54
should be such that the expression Y

which condition will be attained when
is
maximum
(17)
^fe) =
/fe)
*
QJ
^v vî/
"^
v^z/
tt
01
uj(X
/fe)J
/fe)
+ ^log/(o:n = 0.
)]
dq
Now let d logf(x) =
<j>(x)dx,
where
is
<f>
function of x, thus simply related to
/.
another unknown
Then
canceling
out the 7,
)=0.
dq
If
dq
(18)
dq
the results of the respective n measurements on q be
by
s2 ,
-",s n
having
then the errors x are (Art. 7)
designated
Si,
definite, fixed values,
THE ERROR EQUATION
55
from which, at once,

dx l
dx n _
dx
==..._
1Q x
i.
îy;
= 0.
(20)
dq
dq
dq
.,
(18) then reduces to

)
We
already understand enough of the law of error dis
know that when the number of observations is

very large, the number of positive errors of given magni
tude about equals the number of negative errors of the
same magnitude, and that therefore the algebraic sum of
tribution to
the errors
is
discussion
the
large, we may
fulfilled by the
number
write,
now remains
(21) the
of
observations
therefore,
as
+*
+ xz+
may
indefinitely
0.
(21)
two equations (20) and

from which the original
to deduce from the
form of the function
function /
is
another condition
errors,
xi
It
Since in our theoretical
approximately zero.
<,
then be obtained.
see that the equations are satisfied
It is not difficult to
if
= Kx
2,
4>(x
where
is
a constant.
n)
= Kx n
A mathematical
proof that this
is
is given in Note A, Appendix, being

omitted here to avoid distracting attention from the
the necessary relation
56
We may
main problem.
write, then,
(22)
<t>(x)=Kx-,
or since
$(x)dx
= d
Kxdx.
log y
e***+*.
or
j
This
is
The
d log
d log y
Integrating,
log f(x)
Kx2 +
y,
(23)
one form of the error equation.

expression may, however, be so modified as to ex
hibit the relation to better advantage.
We
that the larger the error, the less likely
it is
the larger x
is,
the smaller
is y.
a negative quantity.
the constant
more
useful
c,
is the
to occur
Clearly, then,
K must be
|K
and
h2
e
Replacing
by
by
the equation assumes the more usual and
,
form
y
This
have seen
= ce-* 2*2
(24)
most important equation in the theory of
errors,
and should be committed to memory.

28.
Discussion of the Error Equation.
It
interesting to examine equation (24) to see
will
how
be
closely
the law of error thereby expressed agrees with the conclu

sions already reached.
The
bilateral
symmetry
from the occurrence of x
of the function
is
evident
in the second degree only.
indicates the equal probability of positive

errors of the
same magnitude.
The
This
and negative
function approaches
THE ERROR EQUATION

zero as x increases in magnitude
57
which means that very
great errors are extremely improbable.
The
derivatives
of the function are
(25)
= -2ch 2e- h2*2 [l-2h 2x 2].
dx
From
these, since
maximum
-^
dx
value of y
0,
when x =
<
dx 2
when x =
that
is,
(26)
0, there is a
the error zero
has the greatest probability.

The curve shown in Fig. 5 represents the function,
and has some interesting properties. Its symmetry,
asymptotic character
and
central
maxi
mum
merely
trate
what has
been deduced
illus
just
from
The
the equation.
intercept, or
mum
ordinate,
quantity
c,
when x =
maxi
is
the
since y = c
0.
If
FIG. 5
we put
-n equal to
2
dx
zero,
which
is
the
condition for points of inflection, (26) gives

1
-2
k*x*=
0,
(27)
58
This
is
of inflection
by
OD or OD
P and P The
the distance
corresponding to the points
ordinate of these points
is,
substitution,
y<
=+
(28)
-7=,
Ve
and
therefore proportional to
is
The quantity
c represents
c.
the probability of the error
Now the probability of any given error # is a function
zero.
of
both
or
h.
and
It
h,
since
it
if
changes
would thus appear that
we change
either c
and h have something
to do with the precision of the measurements, and that
they are therefore connected with each other.

see later (Art. 54) that this
there
shall
another factor in the probability of a given

depending upon the value of the smallest scale in
terval in terms of
which the measurements are expressed.
The Principle
We
ment
We
the case, and also that
is still
error,
29.
is
are
now
of Least
Squares in its Simplest Form.

make an introductory state
in position to
of the important principle
of science its
name,
which gives
this
branch
the principle of least squares.
fore
we
will
have been stated several times
Be
are through with the theory of errors, the principle

in successively
complicated forms, as the problems to which
become more and more
general.
So
far
it is
more
applied
we have been
considering only the simplest case, namely, that of ob

servations of equal precision
and while
for this case the
probable value
is
upon a single quantity;

method of deducing the most
clear without reference to the principle
THE ERROR EQUATION

of least squares,
how
to observe
still it
59
be interesting and instructive
will
the assumption of the arithmetical
as the most probable value
may
mean
be shown to be in accord
ance with that principle in the simple form here stated.
The simple form
of the principle referred to
is
as follows
The most probable value of a measured quantity that can

from a series of direct observations, made with
be deduced
equal care and
skill, is that
of the residuals is
for which the
sum
of the squares
a minimum.
The law governing the distribution of errors has already

been deduced theoretically, and the experience of number
We have there
less experimenters testifies to its truth.
fore a right to expect that,
of
when we have made
a long series
measurements upon a single quantity, our observations

have grouped themselves around the true value in
will
a manner approximately consistent with the error equation

Then it is logical for us to assume a value for the
(24).
measured quantity, such that the results of the measure

ments will be so grouped with respect to it. This is the
so-called most probable value,
ciple of least squares, in
arriving at
and
any
it is
the office of the prin
case, to point out the
Let the results of the n observations be
Then
if
we
way
of
it.
sn .
designate the most probable value sought
by
m, there
will
PI, pz, --,
pn
arise
Si,
s2 ,
a corresponding series of residuals
each of which
is
found by subtracting
from the corresponding observation s (Art. 7).

be properly chosen, the residuals derived from it
like true errors,
If
m
m
will,
be found to be distributed in accordance
60
with the exponential law of error probability (24), so that

the probabilities of the respective residuals are
The
probability of the simultaneous occurrence of the
assumed system
of residuals is then (Art. 27)
Y = yw
Now
if
yn
= c-e-^ ^+ - -htf).
m, and consequently the residuals
(29)
are to be so
/o,
chosen that the resulting distribution is the most probable

one in accordance with the law of error, these quantities
must be given such values that the probability Y is as

great as possible. But this will be secured, evidently,
2
by making
will
pi
2
/o 2
-f
...
be seen at once from (29)
so chosen that S/o 2 is a
pn
-j.
as small as possible, as
That
to say,
is
minimum, which
is
m should be
the principle of
least squares stated above.
In order to find what this required value of

write
2?
Hence
= fo - m) 2 + (*2 - m) 2 +
= a minimum.
.-
+(s n
is,
we may
- m) 2
dm
or reducing,
which
is
m = Sl + s*+
simply the arithmetical
"
+ Sn
,
(30)
mean of the observations s.
THE ERROR EQUATION
61
EXERCISES
30.
Show how,
1.
in the first experimental exercise of
aim
Art. 11, the errors of
may
many
enumerating as
think of.
can
you
causes,
2.
be due to
many minor
such possible causes as
Find the algebraic sum of the errors of measurement
in the third exercise of Art. 11
also the algebraic
sum
of the residuals.
3.
Plot the curve y
each of the constants

of logarithms
(e
=
c
ce~h2x*, giving the value unity to

and h. This may be done by use
2.718
Let the unit abscissa be
10 squares and the unit ordinate 50 squares.
with the error curves obtained from Exercises

of Art. 11,
Compare
2 and 3
1,
and with the coin problem curve obtained
in
Art. 22.
4.
Draw
a smooth, symmetrical curve which follows
closely as
as
Ex.
3,
Art. 11,
making
law
erties of the
this
possible
the irregular curve obtained

it
in
conform to the known prop
of error as represented in Fig. 5.
From
determine the relative probabilities of the

magnitude 0.l, 0.2, etc., out to 5. By locating
curve,
errors of
the points of inflection, find an approximate numerical
value for
5.
Has
h.
Plot the curve represented

(2
it
.T)
(3
minimum
- xY +
point
(4
by
- xY +
What
(5
x)
(6
does this illustrate ?
x}\
62

The number
6.
of rays in the lower valve of a certain
species of Atlantic mollusk
was counted
in
508 individual
Plot a curve in which abscissas represent the
number of rays
Of
cases.
these,
1
had 14
rays,
8 had 15 rays,
63 had 16 rays,
154 had 17 rays,
164 had 18 rays,
96 had 19
20 had 20
2 had 21
rays,
rays,
rays.
and ordinates the corresponding number of individuals.

What is the probability that two of these mollusks,
picked up at random, will each have exactly fifteen rays ?
(Data from Davenport,
7.
Tests were
Statistical
made on
ascertain strength of grip.
Manual
of Mental
fifty
Methods.)
schoolboys of equal age to
The
and Physical
following data (Whipple,

Tests) are in
hundreds of
grams.
158
210
248
296
348
175
220
262
301
350
193
225
262
310
353
197
225
267
313
375
197
225
269
315
375
403
200
226
270
320
205
235
273
323
430
206
244
280
325
440
208
244
290
330
440
210
245
294
346
508
THE ERROR EQUATION
63
Arrange a suitable curve showing departures from the

average or normal strength from these data.
About two hundred individuals were tested at the

University of Iowa for accuracy of tone perception, the
8.
results being expressed
by the number
of vibrations in
the departure from the true tone (international A, 435

per sec.) that the individual could distinguish.
The data
are expressed in per cent.
PER CENT.
DEPARTURE, VIB.
1
13.8
24.0
25.5
17.3
7.3
12
3.2
17
1.6
23
2.7
30 or over
4.6
Plot a curve representing this distribution, and discuss
its
form.
9.
Out of a
class of exactly 100 college freshmen, the
was
2 was 22,
16,
12 was 17,
was
23,
31 was 18,
was
24,
22 was
was
25,
was
26.
19,
18 was 20,
12
was
21,
Plot curve and discuss
its
form.
age
64
10.

Out
of over 100,000 public school

grades
by Mr. L. L. Fishwild,
1
per cent, were
50,
per cent, were
55,
2 per cent, were
60,
2 per cent, were
65,
5 per cent, were
70,
6 per cent, were
75,
13 per cent, were
80,
13 per cent, were
85,
25 per cent, were

23 per cent, were
90,
95,
9 per cent, were 100.

Plot curve and discuss
its
form.
examined
CHAPTER V
ON THE ADJUSTMENT OF INDIRECT
OBSERVATIONS
31.
Observations on Functions of a Single Quantity.
has been pointed out that measurements are seldom

made directly upon the quantities whose values are
It
sought, but are usually
functions
consider
As a
it
functions of them, or
them with other unknown quan

case
being
the
simpler,
we
shall
first.
problem,
specific
made
be
involving
The former
tities.
made upon
let
number
upon the diameter of a
circle,
of
measurements
with the object of
determining its area. That is, the quantity really sought

is the area, but the direct measurements are made upon the
diameter, a function of the area.
tions to be
arises,
what
made
all
is
in the
Supposing the observa
same manner, the question
the most probable value of the area
the arithmetical
mean
of the areas
Is
it
computed from the
separate measurements on the diameter, or
is it
the area
determined by taking, as the diameter, the mean of the

measurements upon it? The two are of course not the
same.
This question
deduction.
p
may be answered by the following general

The quantity whose most probable value is
65
66
sought being
q,
and the function
servations s are directly
of
it,
upon which the ob

/(g), there arise the
made, being
following approximate statements, known as observation

equations, each of which represents one of the n measure
ments:
-,
= *i,
f(q)
(31)
/(<?)=
Si,
82,
scale or
*n-
sn are the results of readings
on some
sort of
measuring instrument applied to the function
directly measured.
The
Si
errors
f(q), etc.,
of
the
observations
are
but are not determinate.
we
probable value of q that
are seeking,
and
represented
by
It is the most
if
this
be repre
sented by m, the residuals of the n observations are

Pi
pz
=
=
-/(m),
-/(ra),
(32)
There
is
no reason
why
- f(m).
the principle of least squares
should not apply to this case as well as to the case of direct

measurements, since the law of error distribution, or the
"
law of
departures," is
to this sort of
universal in
scope.
As
relating
measurements, then, the principle of least
squares takes the following form
an unknown quantity
The most probable
can be derived from a

of observations upon one of its functions is that for which
value of
set
its
that
ADJUSTMENT OF OBSERVATIONS
sum
the
of the squares of the residuals arising
The sum above

2p
in
from
these
minimum.
observations is a
67
[si
f(m)]
which
m may
referred to
is
f(m)]
-\-[s 2
expressed
...
-\-
by
-\-[sn
/(m)]
2
,
(33)
be regarded as a variable whose value
to be so adjusted as to render
2/3
minimum.
is
This con
dition requires that
dm
or, differentiating (33),
2 [S*
nf(m}] -fdm
0,
(34)
Therefore
ra,
the most probable value of
whose /-function
is
the
mean
q, is
that value
of the observations
Thus, the most probable value of the area of a

as determined from measurements
is
measurements.
circle,
upon the diameter,
times the square of the arithmetical
of those
upon
mean of the
results
A multitude of other illustrations
of this principle will occur to
any one familiar with such
work.
32.
known
Observation Equations for

Quantity.
More Than One Un
Very frequently,
in
an experimental
research, occasion arises to determine, not merely one, but
68
several,
unknown
quantities or
constants which are so
involved with each other and with the phenomena directly

observed as to render their separate measurement im
possible.
The
following illustrations will
make
this clear.
In the use of the zenith telescope for finding the latitude

of a station, the quantities first sought are the zenith dis
tances of two stars selected for the purpose. The sum of
the zenith distances is equal to their difference in declina
and therefore depends

many very precise measurements
tion, as given in the star catalogues,
upon the
results
of
made with other instruments at fixed observatories.

The difference of the zenith distances is measured by means
of
the micrometer belonging to the
as the instrument
the vertical.
is
zenith
telescope,
rotated from north to south about
In this way, neither zenith distance
is
separately determined, both being found by the simulta

neous solution of the equations arising from the above
observations.
Again,
it is
desired to find the relative proportions of
sodium chloride (NaCl) and potassium chloride (KC1)

in a mixture of the two salts.
Or specifically, in a given
specimen of the dry mixture, to find the number of grams,
x, of sodium chloride and the number, y, of potassium
chloride.
*i.
First let the sample be weighed, with the result
Then
The sample
x
is
now
+y
dissolved
SL
and the chlorine precipitated
with silver nitrate (AgNOs), and the total amount of

chlorine present calculated
by weighing the
precipitated
Denote the
silver chloride (AgCl).
Now,
sodium chloride
chloride
is
is
chlorine
is
of
.6123 x
total chlorine
by
#2 .
0.6123 chlorine, and potassium
Hence
0.4754 chlorine.
chloride and y grams
69
x grams of sodium
in
potassium chloride, the total
+ 0.4754 y =
s2 ,
which furnishes the second observation equation necessary

This is another instance in which
for obtaining x and y.
neither of the
unknown
quantities
is
measured separately.
unknown
Quite often only certain ones of the
quantities
are really desired, the others being merely troublesome

corrections or instrumental constants which
termined or eliminated.
is
ever,
33.
The method
must be de
of procedure,
how
the same in this as in any other case.
More
Observations
than
Quantities.
Normal
In the illustrations of the preceding article

Equations.
there were, in each case, two unknowns, and two inde
pendent observations were necessary to determine them.
By
independent
made on
observations
are
meant observations
a different principle, or under such different con
ditions that the resulting observation equations will

different
terms.
coefficients
To
and not merely
have
different absolute
repeat the process of measuring the
sum
of
two unknowns, without attempting to find some other

relation between them (as, for example, their difference
or their product),
would give no information as to the
separate values of the unknowns.
termination of
of
unknown
And,
in general, the
de
quantities requires a knowledge
independent and consistent
relations
between them.
70
If

measurements could be made without
error, the solu
tion of the I independent observation equations formed

from such measurements would give us the values of the /
unknowns exactly; more than

superfluous.
But, as in the simpler case of a single un
known, the existence

able
to get as
of accidental errors
many
some means
devise
measurements would be
makes
is
desir
observations as possible, and to
of averaging
them
so as to find the
most probable value of each of the unknowns.
lem
it
This prob
the most important that arises in least squares.
Let there be n observations upon

known connected quantities q\, #2,
functions of the
,
qi
(n>l),
un
and
let
the series of resulting observation equations be repre
sented by
/i (qi,
$1,
,
qi)
(35)
fn (qi,
qz,
",
qi)
=s n
Here, as in the simpler cases, there are errors and residuals
obeying the same law of error distribution set forth in
We
the error equation.
probable values, mi,
unknowable)
furnish,
are seeking to obtain the
ra 2 ,
quantities
z,
that
of
the
the
observations
and when these are found, the n
given by
Pi
P2
= si-fi
= #2 ~/2
(mi, mz,
m,),
(mi, W2,
mi),
most
unknown (and
will
residuals will be
(36)
Pn
- fn (mi,
TO,)..
The
principle of least squares
wording to
fied in
values of
unknown
fit
may now
this case, thus
be slightly modi
The most probable
quantities connected by observation equa
which will render the
tions are those
71
of the residuals arising
from
sum
of the squares
the observation equations
minimum.
through an application of
It is possible,
to reduce the
may
number
residual equations (36) to a
number
I,
unknowns, which can then be

the most probable values m. The process
equal to the
solved for
this principle,
of
be regarded as finding from the n observation equa

I most
probable equations whose solution
tions (35) a set of

will give the
From
2
-\~Pn
,
mi
first
most probable values of the unknowns
the principle of least squares, the
must be a minimum, and
may
be so selected that this
2
/Oi
q.
2
/o 2
in order that mi,

will
2,
be the case, the
2
partial derivative of this 2/o with respect to each
of those quantities
That
sum
must be
zero.
(See
any
calculus.)
is,
= 0,
d mi
d
(37)
n
IE, [s
The equations
- f (mi, m
m )] = 0.
2
2,
(37) resulting
from these differentiations
are the most probable or normal equations required, and
72
being
in
number,
values mi,
2,
Equation (34)
one unknown, m.
34.
Reduction
is
solution, the
most probable
which are sought.

a normal equation, containing only
z,
is
of
In nearly
Degree.
on
will yield,
Observation Equations of the First

all
cases in which the
method
of least
used in the reduction of observations in accord
squares
ance with the foregoing theory, the observation equations
are either
the
all of
be
substitutions,
first
degree, or they
suitable
observation
by equivalent
The mathemati
degree.
replaced
equations which are of the
may, by
first
normal equations
are then comparatively simple, and can be performed with
cal operations required in finding the
out any knowledge of calculus.
Let the n
first
degree observation equations upon the
quantities q (corresponding to (35)) be as follows
=
=
+ 6292
2
The
si,
s2 ,
(38)
+ cn q
-----\- r n t
q
3 -\
s.
residuals will then be

Pi
si
- (oirai +
6im 2
+
(39)
Pn
sn
Only one term
in each of these expressions contains
mi
denote the balance of the expression in each case by a single
letter,
as B.
73
Then
pl
= - aimi + Bi,
etc.,
and
- a nm
Differentiating with respect to mi, as per equations (37),
am
Or dividing by 2 and remembering that

in each term,
This
is
...
a np n
+#
= 0.
(40)
the normal equation pertaining to mi, and corre
sponds to the first of equations (37).

This result may be directly obtained by multiplying
each of the residuals (39) by the coefficient of mi in the ex
pression for that residual, adding the results and equating
sum to zero.
The remainder
the
of the
normal equations required are
determined with respect to mz,
ra 3 ,
mj in the same
manner.
The
foregoing processes
following rule
To adjust a
may
be
summed up
set of observation
in the
equations of
the first degree, write the expression for the residual corre
sponding
to
each observation equation, multiply
it
by the
unknown, in that expression, add the

and
their
sum to zero. The result is the
equate
products
normal equation pertaining to the said first unknown.
coefficient of the first
Do
likewise for each of the other
unknowns.
Then
solve
74
the
normal equations thus formed for
probable values, mi, mz,
the desired most
m/.
Let the student prove that taking the arithmetical mean

of a
is
number
of direct observations
upon a
merely a special application of this

35.
Illustrations
from Physics.
single quantity
rule.
It will
be of material
assistance to the student to have presented at this point
a number of actual examples illustrating the application

of least
square adjustment in various departments of
These examples are not

made up for
the purpose; they are drawn from actual experimental
notes on research or field work.
"
exact science.
1.
Bridge Wire.
resistance of a
It
"
was desired to measure the
total
Wheatstone bridge wire and at the same
time to calibrate
it,
by comparison with a standardized

The unknown (and unessential)
bridge of another type.
resistance of the connections
and eliminated.
had
also to be reckoned with
The wire was 100 cm.
long,
and the meas
urement was conducted by observing the resistance of the

10 cm., then of the first 20 cm., etc., and finally of the
first
whole wire, the connections entering each time as a con

stant term in the observed resistance. The results follow.
75
Let x be the total resistance of the bridge wire, and
first of
which
is
to be obtained with
all
that
possible precision,
Mathe
the second to be eliminated, as a mere correction.
The observation
matically they are equally important.
equations are
These are the two unknowns, the
of the connections.
+c
0.2 x + c
0.3 x + c
0.4 x + c
0.5 x + c
0.6 x + c
0.7 x + c
0.8 x + c
0.9 x + c
l.Oa + c
.1
=
=
=
=
=
=
=
=
=
=
0.116,
0.205,
0.295,
0.388,
0.503,
0.595,
0.675,
0.760,
0.850,
0.926.
In practice we need not take the trouble to change

symbols in distinguishing between the true and most prob
able values of the
now
unknown
"
("
and
"
")
If
x and
represent the most probable values sought, the
residual
is pi
0.116
(0.1
c), etc.
first
Let the student
apply the rule developed in the preceding article to obtain

the two normal equations, which he will find to be
3.85 x
5.5
+
+
5.5 c
10 c
=
=
3.686,
5.313,
the solution of which gives
x
c
=
=
0.926 ohms,
0.022 ohms.
76
Let the student
any two
tions and solve them for x and
select
of the observation
c,
comparing the
The accompanying
with these most probable ones.
equa
results
figure
FIG. 6
shows the plotted observations, together with the straight

7
line
0.926
100
0.022
R,
upon which they all should lie were there no errors in the
measurements nor irregularities in the wire itself. The
departures of the plotted points from this most probable
line represent the residuals of the ten observations.
2.
The general theory of the
Balance Constants.
arm balance
is
somewhat complicated, but
equal-
in the equation
used to express the sensibility in terms of the load, the

various instrumental constants may all be involved in
two quantities a and
6,
the equation being
bw
=-
Here
is
the load on either pan (grams) and
sensibility, or
77
the gram
one thousand times the deflection produced

The constants a
by a milligram weight laid on one pan.

and
The
b are to
be estimated from the following observations.
observation equations are then
06 =
10 b =
+
a +
a +
a +
a +
a +
a +
a +
a +
a
The adjustment
20 b
30 b
40
50 b
75 6
100 b
125 b
of these
=
=
=
=
=
=
=
11
1
1
1
1
1
1
1
2212,
2265,
2320,
2343,
2316,
2389,
2449,
2563,
2590.
by the foregoing method
the most probable values sought,
a
b
= + 0.0004466,
= - 0.000000518.
Let the student perform this reduction.
gives as
78
36.
Illustrations
from Chemistry.
Volumetric Solutions.
1.
and
acid
It is desired to test certain
solutions to
alkaline
be used in volumetric
chemical analysis, in order to ascertain their exact strengths.
Two common
tions,
may
these.
If
reagents, in the form of tenth-normal solu
be tested
first,
then others
may
be compared to
the two reagents chosen be hydrochloric acid
and potassium hydroxide, the following procedure may

be employed.
quantity of each solution
is
placed in an accurately
graduated burette, the two burettes being supported

side by side.
A small amount (say about 0.2 g.) of finely
pure calcium carbonate
pulverized
(chalk,
CaCOs) is
and
carefully weighed, placed in a white porcelain dish
treated with an excess (say about 50 cc.) of the
HC1
solu
amount being accurately ob

The chalk dissolves and neutralizes part of the
tion from the burette, the
served.
acid, the
set
CO
gas escaping.
under the
KOH
The
burette, and
just
allowed to flow into
line solution
neutral, this point being
porcelain dish
it
enough
is
now
of the alka
to render
it
exactly
determined by a drop or two of
methyl orange or other sensitive indicator previously added

to the mixture in the dish. The amount of KOH solution
thus used
is
neutralized
by the
Part of the acid
also carefully noted.
CaCOs and
The chemical equations
the remainder by the
two reactions
representing the
are as follows:
72.36
(I)
HC1
99.32
+ CaCO
= CaCl
is
KOH.
+ CO + H O,
2
(II)
The
36.18
55.70
HC1
+ KOH =
79
+ H O.
KC1
small numbers above the symbols are obtained from
the molecular weights, and represent the relative weights

of the substances engaging in the reaction.
Let
qi
q2
=
=
a
a.
From
volume HC1
sol.
sol.
Unknown.
used.
vol.
HC1 sol.
neutralized
by CaCOs.
vol.
HC1 sol.
neutralized
by KOH.
vol.
KOH sol. used in neutralization.
aqi
wt.
HC1 neutralized by CaCOs.
a}qi
wt.
HC1 neutralized by KOH.
bq2
wt.
KOH used.
72.36
aqi
0.73
(I)
or
99.32
0.73,
c.
(II)
(a
or
From
KOH in 1 cc. KOH sol.
total
HC1
1 cc.
CaCOs powder used.
aqi
From
wt.
in
wt.
Then
HC1
a =
b =
c =
(a
wt.
a)qi
aqi
these
bq2
36.18
aqi
55.70
aqi = 0.65 bq2
two equations a
giving finally
0.65 bq2
is
0.65,
eliminated
0.73
c.
by
addition,
80
This
an observation equation, the quantities
is
a, b, c
having been measured, and gi, q 2 being the two unknowns

and a series of such experiments (at least two) will yield
;
the most probable values required.
periments the
c
giving
CaCOs powder may be omitted

all of them.
(Why ?)
entirely,
but not in
The above data

tions
In some of the ex
yield the following observation equa
50
50
0i
qi
11.23gi
11.25gi
11.25
0i
11.34gi
0.65
0.65
0.65
0.65
0.65
0.65
X
X
X
X
X
X
0.73
0.73
9.9802
0,
10.0002
0,
10.00
0,
0.
10.33
7.88
10.10^2
X
X
0.1779,
0.1936,
Let the student reduce these to normal equations and solve

for the
2.
most probable values
Pyknometer Constants.
nometer
of q\
and
2.
The expansion
(specific gravity bottle), like
any
of a
pyk-
solid, is in
ap-
81
proximate accordance with the linear law
V=
V being the
and
zero
Fo
+ Kt,
capacity at temperature
Ka
t,
the capacity at
constant involving the coefficient of expan
The two
sion of the glass.
constants
and
must be
experimentally determined from time to time for any

pyknometer that is used in accurate measurements of
density.
This
may
be done by finding the capacity at
several different temperatures over the required range.*
The
following
using distilled
is
a tabulation of eight such determinations,
water and corrected for buoyancy of the
air.
Let the student form the eight observation equations and

the two normal equations, and reduce for the most prob
able values of
Fo =
37.
1.
25.2509, K
and K.
Illustrations
Locating
writers
(The approximate answers
are,
0.0005244.)
from Surveying.
Distant
Station.
Some
of
the
on surveying strongly recommend the use of
best
rec-
* If the range be large,

will vary somewhat.
The range may be
subdivided, say into 10-degree intervals, and the constants found for
each or better, a quadratic relation assumed, with three constants. See
Art. 45.
;
82
tangular coordinates in surveying and mapping

their use reduces
basis.
many
The problem
in
calculations to a
hand
is
more
as follows:
certainly
scientific
Given, the
FIG. 7
reference to
station
number
of stations A, B, C, etc., with

the bearing of an unknown
and
origin 0,
from each of these stations; to find the most
coordinates of a
an
probable coordinates of P.
For instance, the unknown
station
83
a reef near the coast along which the points
is
The numerical data
A, B, etc., are located.
are as follows,
for five stations.
The vectorial angles in the last column are the angles

made by the vectors PA, PB, etc., with the line drawn
eastward through P, calculated from the given bearings.
Using coordinates x and y to locate P, and x a y a etc., for
,
A,
we can
etc.,
write
y a -y
x
xa
or
that
x tan
Ba
tan
x a tan
1785 tan 31 3
fl
6a
ya
is,
x tan 31
etc., as
1501,
the observation equations, there being as
known
of
These equations, being

be
y, may
adjusted in the usual
Let the student do this. (The results should
these as there are
be, approximately, x
E., 1000 N.)
stations.
and
of the first degree in x
manner.
many
930, y
1000
that
is,
P =
930
84
2.
is
taken from Merriman
many
is
The next
Relative Levels of Stations.
Least Squares, and
illustration
is
typical of
kinds of measurements in which the quantity sought
measured by parts or segments.
for example, applied to a
number
The same method
of angles at
is,
one station.
Given, a number of determinations on the relative altitudes

of several stations, obtained
by
precise leveling,
to find
the most probable values of their altitudes above one of
them taken
as a datum.
Following are the results of the
levelings.
A
B
B
above
573.08
ft.
2.60
ft.
above
575.27
ft.
C above B
167.33
ft.
above
D above C 3.80 ft.

D above B 170.28 ft.
D above E 425.00 ft.
E above
E above
Representing by
stations
above
319.91
ft.
(one way)
319.75
ft.
(another way)
a, b, etc.,
the elevations of the respective
as a datum, the following simple observa
tion equations at once result.
=
b - a =
b =
c - b =
d-c =
a
573.08
2.60
575.27
167.33
3.80
-b =
d-e =
e
c
The student can
=
=
170.28
425.00
319.91
319.75
readily adjust these in the usual manner.
It will be interesting in this case to
values of
and
compare the adjusted
with their values as directly measured.
from Astronomy.
Errors of the Transit and Clock.
1.
ascertained, at
the stars.
veyor
on a
Illustrations
38.
is
a, b
To
Astronomical time
any observatory, by observations upon

end an instrument not unlike a sur
this
s transit is
solid pier,
85
used.
and
is
It is larger, however,
and
fixed
incapable of rotating horizontally,
being swung in the vertical plane of the meridian. This

instrument is the astronomical transit or the meridian
circle.
When used for time
observations, the telescope
is
set at
the proper angle of altitude for some star to traverse

field as it crosses
the meridian.
The exact
is known
its
sidereal time
as the right
of meridian passage, or transit,
*
ascension of the star, and is given in the star catalogues.
In order to correct the clock, therefore,

to note at
what time, by the
it is
necessary only
clock, the star is actually
ob
served to cross the meridian.
is
* Right ascension on the celestial sphere, as shown by the star maps,

on the earth, only it is usually expressed
closely analogous to longitude
minutes and seconds, reading toward the east.

corresponds to terrestrial latitude.
in hours,
Declination
86
On
account of the extreme accuracy demanded in as

tronomical work, this apparently simple procedure re
quires the elimination of certain recognized instrumental
These are
arising
from the
non-horizontality of the bearings or trunnions
on which
errors.
(1)
the
the telescope turns, so that
level error,
its
revolution does not exactly
coincide with the meridian plane
(2)
the azimuth errort
or failure of this axis of rotation to coincide with the east
and west
line,
of rotation
(3)
which has a similar
effect
on the plane
the collimation error, due to the fact that
the cross-wires in the telescope, which determine
its line
of sight, are not exactly in the optic axis, being a little to
one side of the center of the

there
field.
the error of the clock, which
is
wanted.
In addition to these,
is
the quantity really
The level error is ascertained by a
tion of the stride level resting
direct applica
on the trunnions and having
a very sensitive graduated spirit-bubble.
The
other errors
must be found simultaneously from several observations

on different stars, the level error reading being simply a
part of the determination.
Without entering into the applications
of spherical as
tronomy required,
may be simply stated that the ob
servation equations involved are of the first degree. If
it
=
<?i
the azimuth error,
q2
#3
I
all
the true clock error (clock minus true time),
=
=
the collimation error,

the level error,
being expressed in seconds of time, then the form of
the observation equation
qi -f
ag2
87
is
d being the apparent clock
cg 3
bl,
time indicated by
the clock at apparent transit minus the true time of transit,
The quantities a, b and c
or right ascension, of the star.
are
known
as
error, or the
s coefficients,
Meyer
the following formulas
sin
(X-a)
cos
b
and are calculated from
= cos(X-S)
cos 5
in wilich
is
sec
the latitude of the observatory and
lination of the star used.
at
hand
5,
the dec
Tables of these coefficients are
in every observatory.
Of course three observations on
different stars, at least,
are required to determine q\ and eliminate qz
more are made,
q%.
If
may be applied
a typical set of data
to their adjustment.
of this sort,
and
least-square reduction
Following is
based on the observed transits of
IOWA CITY, IOWA,
Lat. 41
40
November
six stars.
16,
1896
88
No
allowance has here been
clock
made
Parallax and Proper Motion.
2.
for
error in the
any
rate during the progress of the observations.
Stellar parallax
is
the
in the position of a star, during the year,
apparent change
caused by the earth
this, there
is
motion
the actual, or
In addition to
in its orbit.
"
proper,"
motion
of the star
through space. These two are superposed and pro

duce one resultant effect upon the star s apparent posi
itself
tion at
any time.
components
Modern
is
Their separation into distinguishable
the problem here presented.
astronomical
measurements
very largely by photography.
The
conducted
are
star in question
is
photographed on the same plate with others so immensely

farther away that they have no perceptible parallax or
proper motion, and then the positions of the images are
measured at leisure on very accurate measuring machines.
Let
TT
/*
=
=
=
the parallax in a given direction,

the proper motion in that direction,
the measured displacement of the star in that
direction,
with reference to
position at
some previous date T days
Then the observation equation

astronomy to be
PTT
is
7>
+c=
apparent
past.
in practical
s.
the parallax factor, easily calculated from the direc-
tion of the star

c is
shown
is
its
and the position
of the earth in its orbit.
an unknown constant, depending on the peculiarities

measuring machine, and to be eliminated. The
of the
three
unknowns
are, then,
TT, f^
and
c.
The
89
coefficients
and T and the quantity s are varied by making observa

tions on many different dates, and from the resulting
series of
of
observation equations, the most probable values
the proper motion and the parallax are obtained.
latter gives the
The
The
most probable distance of the star.

somewhat technical, no numeri
details of the process being

cal
example
is
here given.
If
Observation Equations Not of First Degree.
are
not
of
the
first
re
degree,
39.
course
may
be had to the general method explained in
Art. 33, that
is,
to the application of the principle of least
squares through the general equations (37).
This would
often lead, however, to normal equations that would be
exceedingly inconvenient to solve.
In
many such cases, the difficulty may be at once avoided
by a suitable application of logarithms. A standard

measurement in the physical laboratory, for example, is
the simultaneous determination of the magnetic field of
the earth
magnet used
and the magnetic moment

for the purpose.
of the bar
One experiment
gives the
product,
(41)
SI>
and another the quotient

(42)
f-*,
of the
may
unknown
quantities.
These observation equations

and H, as
be made linear by using instead of
90
unknowns, their common logarithms
M + log H = log
(43)
- log H = log
M
log
the most probable values of log M and log H being then
log
$1,
s2 ,
found
in the usual
manner.
Again, the solubility of a chemical salt

theoretical
formula*
= se
(44)
terms of the centigrade temperature

C. and c
solubility of the salt at
ing on
given by the
ct
in
is
heat of solution.
and
is
t.
is
the
a constant depend
unknowns, to be
determined for each substance by means of several meas
urements on s at different temperatures. For this purpose
the observation equation may be written
its
c are
log s
273
log e
the most probable values of c and log s

directly found.
bility of
The
log
s,
(45)
being the values
following data pertain to the solu
potassium chlorate (KClOs) in water.
* See Arrhenius, Electrochemistry.
The
eight observations
on t and
s furnish eight
91
observation
equations of the above form, which when adjusted give

as most probable values log S Q =
1.6073, whence SQ =
0.0247
and
substance
The
13.82.
may now be
solubility formula for this
written in
its original
form, or more
conveniently retained in the logarithmic form

log s
6.0014
from which the values of
The student
calculated.
the observations on
and
i-
1.6073,
given in the third column are
will find it instructive to plot

s,
and
also the
corresponding to the calculated values of
smooth curve
s.
It
would be
imagine a more typical application of least-
difficult to
square adjustment than the one just given.
Another method
of procedure
equations are not of the

to Horner s
is
equations,
40.
method
of
first
when the observation
degree,
somewhat analogous
approximation for higher algebraic
explained in Note B, Appendix.
Observations upon Quantities Subject to Rigorous
Conditions.
It often
happens that unknown quantities
involved in observation equations are further connected
by known mathematical conditions, which the final ad

justed values must rigorously satisfy. For example, the
most probable values of the angles of a triangle could not
be a set of angles whose sum is other than exactly 180
;
the
sum
of all the percentages in a chemical analysis
must be 100;
are
known
etc.
Observations upon such quantities
as conditioned observations.
92
Suppose that the
results of
three angles of a triangle are
(46)
These are the observation equations. To

must be added a fourth equation, namely
this list there
qi
which
is
called
?2
q*
= 180,
(47)
an equation of condition.
It differs
from
the others are not.
known to be exactly true, while

The three most probable values,
when deduced, must
satisfy this equation exactly;
the others in that
others
must be
it is
satisfied as nearly as
may
be.
the
This equa
tion of condition cannot, therefore, be classed as an ob
servation equation and treated like the others.
In general, we may have n observations involving I un

knowns, which are further subject to m rigorous conditions,
expressed as equations of condition,
for
if
equal to
it,
m must be less than
the unknowns would be absolutely de
termined by the given conditions, and the measurements

would be superfluous and if greater, no set of quantities
;
could, in general, be found to satisfy
all
the conditions.
There being fewer conditions than unknowns, there is

an unlimited number of sets of values of the unknowns
which might satisfy the conditions, and we have to de
termine from the n observations which of these sets is
the most probable.
93
m conditions do not give the values of the

they enable us to express m of the unknowns
Though the
unknowns,
we now
unknowns in the
may
then be adjusted
rigorously in terms of the remaining ones;

substitute these expressions for the
observation equations, the latter
most probable values of the I

maining. The most probable values
for the
quantities
may now
and
if
quantities re
of the
replaced
also be calculated so that the condi
tions are exactly satisfied.
Applying this to the case of the angles of a triangle,

subject to one condition (47), one of the angles, say #3,
be expressed by means of
may
tW
93
The
180
it
9l
in
terms of the other
?2 .
(48)
three observation equations then appear:
(49)
180
Let the student adjust these and show that the most
probable values sought are
[180
(50)
the third result folio wing, from the other two through sub
stitution in (48)
which shows that the
results sought
can be obtained by adding to each measured angle onethird the discrepancy between the sum of the measured
;
94
angles and 180, so as to
make
on the assumption that
all
are equally trustworthy.
proceeding
is
the
sum
correct.
This
is
three of the measurements
(See Chap. VII.)
The same
to be followed in every case where the
observation equations represent the separately measured

values of the unknowns, while the one equation of condi
tion rigorously gives their sum.
common
If
Cases
like this are of
occurrence.
the sides of the triangle are measured, as well as the
angles, there will be six observation equations (at least),
and three equations of condition. Of these latter, one

will be the same as (47), the other two arising from the
requirements of trigonometry as to sides and angles.
A case of special importance to the surveyor is the ad
justment of the sides and angles of a polygon of land. In

addition to whatever measurements are made upon the
lengths and bearings of the sides, there are two rigorous
conditions to be fulfilled, namely that the algebraic sum of
:
the projections of the sides on an east-and-west line
and the algebraic sum

south line
is
zero.
in detail in the
of their projections
This adjustment
will
is
zero,
on a north-and-
be found explained
more advanced works on plane surveying.

EXERCISES
41.
1.
Draw
a large triangle on paper with a fine pencil,
and measure with a protractor each
of the angles.
Form
the observation equations and the equation of condition,
and from them deduce the most probable values

angles.
of the
2.
Lay
off
on a straight
line four points,
95
A, B, C, D.
Measure AB, BC, CD, AC, BD, AD. From these

measurements form observation equations and compute
the most probable values of AB, BC, CD. These seg
ments may be conveniently
3.
The
lettered x, y,
following measurements were
z.
made upon
tangular metallic tank to determine its dimensions
Length
(inside) 27.31
a rec
:
cm.
Width
(inside) 16.08
Depth
(inside) 9.67 cm.
cm.
Capacity by standard graduates, 4.3217
liters.
Find the most probable dimensions.

4.
Draw
five lines radiating accurately
from a common
point 0, the further extremities being A, B, C, D, E.
Measure with a
protractor, by the differential method,

and turning the protractor at each measurement, each of
AOB, AOC, AOD, AOE, BOC, BOD, BOE,

COD, COE, DOE. Determine the most probable values
of the angles AOB, BOC, COD, DOE.
the angles
5.
The
following are the results of an analysis of a cer
tain medicinal
compound
Salts of calcium 1.26 per cent.

Salts of
sodium 2.53 per
cent.
Salts of iron 0.23 per cent.

Salts of
manganese 0.14 per
cent.
Salts of quinine 0.07 per cent.
96
Salts of strychnine 0.02 per cent.
Water 95.67 per
cent.
Find the most probable values

6.
Six points, supposed to
of the several percentages.
lie
on the arc
the following measured coordinates
of a circle,
have
Find the most probable coordinates of the center and
most probable
7.
radius.
A steel tape was
of stretch
measured under different conditions
and temperature, as follows
Using the approximate formula

bf,
in
which
temperature and /
the most probable values of
IQ,
a,
=
b.
tension,
adjust for
8.
Let
(Adapted from Wright
Adjustment of Observations.)
be the difference in length of two standard meter
bars at 62 F. and
Then
of expansion.
temperature
t is
the difference in their coefficients
the difference d in length at any
rf
Observations were
made
_
(<
62)
as follows
Van
24.7
0.00791 inch
37.1
811 inch
61.7
833 inch
49.3
820 inch
66.8
847 inch
71.5
849 inch
Adjust for the most probable values of

9.
97
der
Waal
and A.
equation for pressure and volume
T may be put
of a gas at absolute temperature
v*TR
va
The measurements
of
Amagat on
+ pv*b
-f ab
air at
in the
form
pv*.
moderate pressures
and at 16 C. (289. 1 absolute) were published as follows

p IN CM.
MERCURY
pv
76
1.0000
2000
0.9930
2500
.9919
3000
.9908
3500
.9899
4000
.9896
98
Form
Appendix, and adjust

10.
The
by the method
electrical conductivity of
selenium
vary with the intensity of light falling on

the equation
The
of
Note B,
for a, b, R.
C=
it
is
found to
according to
following data were furnished
by Dr. F. C. Brown.
83
3
188
11
250
17
285
25
303
Adjust for the most probable values of a and b. (NOTE.

In working the above problem, it will be found
necessary, as is sometimes the case, to use caution in
dropping decimal places, as the normal equations hap
to slight changes in the
sensitive
pen to be quite
"
"
coefficients.)
11.
The E.M.F.
perature difference
by
of a thermo-couple for a given

t
between junctions
the equation
e
The
at
bt
may
tem
be represented
2
.
following values for a copper-tellurium couple with
one junction at 0, in which
by Mr. W. E.
Tisdale.
e is in volts,
were furnished
99
<
50
0.000243
150
0.000254
82
242
162
262
245
180
265
248
186
267
113
249
190
266
117
250
195
267
128
252
200
268
140
251
92
100
Calculate the most probable values of the constants a
and
Plot the curve.
b.
In a sine intensity magnetometer, let the pole

strength of the bar magnet be P, the distance between its
12.
poles
a.
and the distance from
/,
d is
its
center to the needle pivot
the needle deflection and
sity of the earth s
these quantities
magnetism.
H the horizontal
inten
The equation connecting
is
-*..
The
following data were obtained at a station where

0.1884(c.g.s.).
a (CM).
a (CM).
20
24
17
40
25
12
46
45
35
30
48
50
27
35
26
Find the most probable values of

method of Note B, Appendix.
and L
49
Use the
100
13.

The
specific
volume
of a certain liquid
was meas
ured at
different temperatures by a quick secondary

method which was known to have certain small persistent
errors.
At three of the temperatures, known as
tie
"
points,"
the specific volume was also measured by a more
laborious absolute method, free from the said sources of

error.
The
results follow:
In order to correct
was made
all
the data in the second column, use
of the equation
Y =
in
which
AX +
B,
X is the specific volume by the secondary method
and Y the corresponding corrected value,

assumed constant. By using the values
and
of
and
being
at
the three
and
14.
"
tie
points,"
and correct
all
find the
was calibrated
of
the secondary data accordingly.
glass sinker, used for precision
of liquid density
101
measurements
by the Archimedes buoyancy method,
for expansion, the
data being as follows
VOL. SINKER
TEMP. C.
24.0
(cc.)
34.03894
3907
24.5
25.0
3984
25.5
4085
26.0
4102
26.5
4134
27.0
4191
27.5
4203
28.0
4231
28.5
4240
29.0
4290
30.0
4393
Find the most probable zero volume and coefficient of ex

pansion of the sinker, assuming a linear relation.
16.
Guthe and Worthing
Iog 10
formula for the vapor pres
sure of water at temperature
p = 7.39992
C.
is
(J
+ 273)
From
the following data, find the most probable values of
a and
b.
102

p (MM. OF MERCURY)
16.
The
10
20
17
30
32
40
55
50
92
60
149
80
355
100
760
angles and the sides of a triangle
measured, with the following
Angle
Side
ABC
were
results.
A
B
51
95
33 51
BC
1721.3
AC
AB
2207.5
ft.
1233.0
Introduce the necessary geometric conditions and adjust

for the most probable values of sides and angles.
17.
its
Draw
accurately a large quadrilateral
two diagonals
AC and BD.
A BCD
and
Measure with a millimeter
AB, BC, CD, DA, and with a protrac

DAB, DAC, CAB, ABC, ABD, DBC,
BCD, BCA, ACD, CDA, CDB, BDA. Introduce the
scale the four sides
tor the angles
eight necessary geometric conditions

rilateral.
of the sides
and adjust
and angles
for the
of the
quad
If
the diagonals had also been measured,
103
how many
conditions would be introduced?

18.
Adjust the transit observations given in Art. 38.
19.
(Adapted
Squares.)
from
Crandall
tions for x, y,
Least
z.
+ 1.41 y + z =
+ 0.68 x + 1.00 y + z =
+ 0.52 x + 1.02 y + z =
+ 2.51 x - 2.67 y + z =
- 0.73 x + 2.13 y + z =
+ 0.75 x + 1.01 y + a =
+ 0.53 x + 1.02 y + z =
+ 0.68 z + 1.00 y + z =
+ 0.81 x + 1.02 y + z =
+ 0.09 z + 1.27 y + z =
20.
and
Geodesy
Adjust the following transit observation equa
0.07 x
0.65,
+ 0.18,
+ 0.13,
+ 3.96,
-
1.88,
+ 0.02,
+ 0.13,
+ 0.44,
+ 0.29,
-
0.76.
In the following observation equations the un
knowns n and
+K
[~1
are constants of metallic reflection.
- } sin
L_
1= sin
TL J"^
^J. ~r~ j\. ^ _J
Assuming that approximate values
of
n and
tan
X are known,
transform these into observation equations of the

degree
by the method
of
Note E, Appendix.
first
CHAPTER
VI
EMPIRICAL FORMULAS
42.
the
Classification of Formulas.
many
If
we examine
*
into
formulas employed to represent natural or
physical laws,
distinct classes,
found that they fall into two fairly

which may be called, respectively, rational
it is
and empirical formulas.
To
the former class belong those which have been de
duced through processes of mathematical reasoning from

the elementary and established laws of the science to
which they pertain; hence the term rational. Such, for

example, are the equations for the motion of falling bodies,
the expressions for electric or gravitational force at a
point, the equations of the balance, the error equations
for the astronomical transit, etc.
certain constants or
is
coefficients,
In these there appear
the determination of which
often a matter of great scientific importance.
Empirical formulas, on the other hand, are those whose
form
from the
inferred wholly
is
results of experiment
or observation, and which have not been deduced theoreti

cally.
Some
of the best
examples of these are to be found
in engineering, such as the formulas for the flow of water

in pipes
and channels, or
of temperature.
for
steam pressure as a function
Empirical formulas also contain con104
EMPIRICAL FORMULAS
slants,
as
if
which are determined
in exactly the
105
same manner
the formulas were rational, and whose determination
depends upon experiment and measurement.

A closer examination into the subject reveals, however,
the fact that the boundary between these two classes is by
no means a sharp one, for the reason that a very large

proportion of the rational formulas purporting to represent
upon more or less em

pirical and approximate assumptions, which have been
adopted for the sake of simplicity of form, or for want of
natural laws have been deduced
better information.
whether there
exist
In
fact,
it
may
be doubted
well
absolutely rational formulas per
any
taining to material magnitudes.
Even Newton
great
and
it is
possible that some future investigation in astronomy
may
law of gravitation has

demonstrate
43.
it
its
experimental basis;
to be inaccurate.
Uses and Limitations
of Empirical Formulas.
Empirical formulas owe their existence to the fact that in
many
deduced to repre
phenomenon, but that,
cases no rational formula can be
sent the law of behavior of a

nevertheless, experiment
shows some law
is
being obeyed
which appears to be simple in character and

presumably expressible, at
matical symbols.
is
therefore
least approximately, in
Not being
able to trace the
operating between cause and effect, on account of

plexity or for other reasons, the experimenter
more or
less blindly for
mathe
mechanism
its
com
must seek
a functional relation that will
satisfactorily connect them.
It
may happen
that the
106
finding of such a relation as accords perfectly with the
observations will throw
much
light
on the nature
of the
mechanism
itself, and lead to a theory relative to it, which

can be tested by more intelligently directed later ex
Stefan
periments.
fourth-power law of cooling, which,
though wholly empirical as
far as Stefan
was concerned,
has led to the important modern theory of radiation,
an excellent example
is
of this sort.
But the great majority of empirical formulas are con

fessedly artificial, and reveal nothing of the real nature of
the connection between the
phenomena
do not even pretend to consistency

sions
in the
involved.
Many
matter of dimen
the writer has estimated railroad culvert openings,
for example,
on the crude working
rule that the area of
opening, in square feet, should be equal to the square root

of the drainage area, in acres
an area equal to a length.
Nevertheless these formulas are capable of the utmost

practical usefulness;
upon the
principle of
by means of them, depending

continuity, we may accurately inter
for
polate the values of the
unknown
function between points
actually observed, and even, in a limited way, extrapolate
beyond the experimental region into conditions unattain

able in practice.
There
is still
another class of empirical formulas, more
or less in the nature of scientific curiosities, which repre

sent, in the experimental region only, a relation
between
variables that have no conceivable connection with each

other.
It
is
mula which
thus possible to construct an

will follow,
with
fair
artificial for
accuracy, the increase
EMPIRICAL FORMULAS
in population of the
United States, or of a
107
city,
with time,
or even the fluctuations of the stock market over a given

interval of time.
Such formulas
however, of
are,
little
value, as they are merely a sort of cast of a series of statis

tics
which are themselves available
represented
may
and
since the variable
not even be continuous, interpolation
and extrapolation with any certainty are impossible.

It must also be pointed out that empirical formulas
cannot be allowed to enter into theoretical developments
on the same basis as rational ones, unless their physical

nature is first carefully looked into and the region in which
they are assumed to apply
Where the
is
properly circumscribed.
(supposing one to
true functional relation
exist) can be dealt with mathematically with safety, an

artificial one closely approximating it may lead, if so used,
to altogether erroneous conclusions.

44.
1.
Illustrations of Empirical Formulas.
Reduction of Pendulum
reversible
pendulum
is
to
The Kater
Zero Arc.
familiar to nearly every physical
laboratory student as a means of obtaining the accelera

tion of a falling body,
g,
or the value of
so adjusted that the time of swing
supports,
i.e.,
points, the
when the
is
gravity."
When
the same from both
knife edges
pendulum swings
"
are
at
in a period given
conjugate
by the
ideal
simple pendulum formula
in
which
is
the distance between the knife edges.
The
108
determination of g
is
therefore a matter of measuring the
period of oscillation and the distance
I.
This equation affords, however, an excellent example of

the class referred to in the last paragraph of Art. 42. For
in its deduction
assumed that the pendulum swings

friction from a perfectly rigid support,
it is
without any kind of
and that the amplitude of vibration is infinitely small,

none of which conditions is attainable. The writer has
attacked these
good
difficulties in
the following manner, with
results.
Apparatus is arranged to release the pendulum so as to

swing with any desired initial angle of amplitude, and
the time accurately observed for each of several small
The
amplitudes.
my
students,
degrees,
factors
are typical.
is
<f>
by one
of
the half amplitude in
the period in seconds.
The steady
all
following results, obtained
increase of period with amplitude includes
the true, theoretical increase that would exist
under ideal conditions, and the influences of

pivot friction
and bending
of
supports.
air friction,
The
results
are plotted in Fig. 8, which shows an unmistakable cur-
EMPIRICAL FORMULAS
vature with downward concavity.
fore
not linear, but
The
empirical formula
may
T = a
is
0.
now assumed to
This
is
relation
is
there
be approximately quadratic.
bcj>
+ ctf
(51)
7
represent the variable T as a function of
treated as a form of observation equation in
FIG.
which the
The
109
coefficients a, b, c are the
servation equations
and
unknowns.
The ob
residuals are written out as usual,
the normal equations deduced and solved, with the ap
proximate results
a
b
c
= + 0.878400,
= + 0.000073,
= - 0.000001,
110
which give
T =
0.878400
as the relation desired.

is
the value of
that
is,
This
is
0.000073
In reality only a
for zero
amplitude that
the limit approached by
0.878400
By
sec.
fore, it is possible to
0.000001
as
is
(52)
<
wanted, for
we
it
are seeking
$ approaches
zero.
this slight extrapolation, there
extend the experiments into a region

The value of g may now be
unattainable otherwise.
calculated from this result and the measured distance
between knife edges.

2. Solubility Formula.
Previous to the theoretical
calculation of a rational formula for solubility in terms of
temperature (Art. 39), the relation was represented by an

empirical formula of simple power terms
:
The data given
in Art.
of the constants a,
bt
+ ct + dt*.
39 will
6, c, d,
suffice for
the determination
a result being obtained which
will fit the observations nearly,
the rational expression.
(53)
if
not quite, as well as
This exercise
is left
to the stu
dent.
Gordon
3.
Formula for Rectangular Columns.
The ul
timate strength of a rectangular column under compres

sion
it is
is
;
found to depend fundamentally upon how slender
specifically,
upon the
ratio of its length to its shorter
For long, slender columns, the

found to be expressed satisfactorily by the
following formula, in which U is the ultimate compressive
transverse dimension.
relation
is
EMPIRICAL FORMULAS
111
strength per square inch of cross section, and
R the ratio
of length to least width.
(54)
a and
b are the empirical constants to
be determined.
The
following data refer to white-oak timber columns or posts,
being expressed in pounds per square inch, and will
serve as a further exercise for the student.
a and
45.
will
b should
The reader
know by what process the form to be
representing the unknown relation between
Choice of Mathematical Expression.
now wish
used,
be about 925 and 0.00091, respectively.
as
variables,
may
to
be arrived
covering this matter.
at.
The
There
is
no general rule
empirical form once being
settled upon, the calculation of the empirical constants

is
a direct process;
but the selection of a mathematical
expression which can be made,

stants, to
problem
fit
by the use
of proper
the facts with sufficient accuracy,
is
con
often a
calling for the exercise of the highest degree of
ingenuity, especially where there
dependent variable.
is
more than one
in
112
The
first
step will probably always be to plot the results
of the observations, or the data to be represented, to a
suitable scale
some
on coordinate paper.
sort of curve, which,
if
at
all
The
result will be
regular, will give
an
idea as to the nature of the variation, and will often sug

gest
an equation through
known
its
resemblance to some well-
locus, such as the straight
line,
parabola, etc.
few general forms have been found especially adaptable.
The equation
y
continued so far as
bx
may
ex 2
dxz H
be necessary,
(55)
may
be used for
curves which are not periodic, nor asymptotic, nor very

irregular.
The number
of
terms to be used
will
be limited
by the fact that the coefficients of powers that

omitted turn out to be negligibly small.
may
be
This form was
used in two of the three examples in the preceding article,

and might probably have been used with some success
in the third.
It
was remarked
in connection
with the
second example of Art. 36 that the volume coefficient of
expansion really varies when carried over a considerable
This might have been allowed for by adding a term

involving the square of t, with a third unknown constant
range.
coefficient, to those used.
used,
it
will
When
be well to apply to
such a form as (55)
it
is
the values of x and y
belonging to five or six of the observed points that seem
most accurately on the curve, as a preliminary

calculation, and determine from them approximate values
to
of
lie
an equal number
of the coefficients
without least squares,
EMPIRICAL FORMULAS
in order to ascertain
It
may
where the
series
113
may safely be stopped.
be found, in this way, that only two or three terms
are necessary, the coefficients
beyond these being
negli
gible.
The
many
following equations are also quite adaptable to
physical phenomena, particularly those involving
variables which approach a limit, or in which
maxima
and minima do not appear.
+ b log (x + K).
x = a + b log (y + K).
ax + by + c = xy.
= a + b log x.
log y
(56)
tion;
is
(56)
(57)
(58)
(59)
asymptotic in the y direction, (57) in the x direc

is
(58)
an equilateral hyperbola, asymptotic
The
both directions.
logarithmic formulas
may
easily
be put into exponential form if desired. The constant

may sometimes be theoretically assigned.
The equation
is
ax
by -f c
more general and includes
to be a special case of
may have maxima

n may be given a
a, b, c will
it.
or
(58).
The
in
x ny
(60)
(54) will also
be seen
curves represented by (60)
minima and points
small integral value, as
of
1,
inflection.
2,
3,
and
be the empirical constants to be determined.
The student
will
assumed values
do well to plot these equations, using
of the constants.
For functions that are apparently periodic, or which
114

in the course of the varia
ups and downs
there may be used a limited number of terms of the
have many
tion,
"
"
trigonometric series
+ b sin nx + c cos nx
+ d sin 2 nx c cos 2 no:
-f/sin 3 nx + # cos 3 nx +
-j-
This
is
a Fourier
s series,
and can be made to
fit
(61)
any curve
with any desired degree of approximation by carrying

The calculation of
it to a sufficient number of terms.
the constants
may become
extremely laborious, and Prof.
A. A. Michelson devised, some years ago, a mechanism
known
as the harmonic analyzer, which will give their
approximate values.
the aid of this machine
By
possible to analyze very complicated
it
is
phenomena, such as
the tides or the variations in terrestrial magnetism, into
harmonic components, and often to reveal their component

causes.
in
But
it is
an altogether
by this means, to express

manner such phenomena as
also possible,
artificial
To
are referred to toward the close of Art. 43.
empirical
formulas of this class applies, particularly, the caution

against treating
them on the same
basis
as
rational
formulas in mathematical analysis.
Very often the problem
of selecting the proper
form
will
giving attention to obvious limiting con
be facilitated by
ditions, such as the fact that
effect is zero
where cause
is
This amounts to making the selection partly

zero, etc.
rational, and only emphasizes the statement that there is
no sharp distinction between rational and empirical ex-
EMPIRICAL FORMULAS
pressions.
After
all
has been
will still find true the
matter
calls for skill
said,
115
however, the student
remark, previously made, that this
and ingenuity
of a high order.
EXERCISES
46.
1.
Experiments were made upon the index of re
fraction of a solution of varying concentration
sodium
light being used.
The
results follow
and density,
:
Express the variation by a suitable empirical formula,

deducing the constants. Would it be safe to infer from
this
2.
formula the index for pure water?
galvanometer attached to a thermo-electric couple
gave the following readings
y,
for
the
corresponding
differences of temperature x:
Prepare suitable empirical formula, deducing constants.
116
The
3.
curve
following are observed positions of points on a
Obtain an equation whose graph

nearly as possible,
4.
air,
and plot
The temperature
The temperature
expressing 6 in
6.
of a heated body, cooling
was taken each minute
being here tabulated
will fit these points as
it.
in
the
for ten minutes, the results
of the air
terms of
was 20.
Deduce an equation
t.
Measurements were made upon the radioactivity
of a deposit of pure thorium at intervals after its forma

tion, as follows
EMPIRICAL FORMULAS
117
Express the relation as an empirical formula,

activity will die out with time.)
The quantity
6.
of
discharge Q,
(Note that
cubic feet per
in
minute, of a 10-inch sewer pipe was found to vary with

the
data
slope
grade)
as
per
the
following
Work
7.
(percentage
out an empirical formula and plot
The means
of
many
it.
observations upon a certain
variable star of short period gave the following variations

of
magnitude
118
Represent
this
variation
as
by
simple
a formula as
possible.
i
8.
The atmospheric
refraction
horizon at various altitudes
by the following
50 F. and normal
R
is
for a star
above the
given approximately
corresponding to temperature
table,
pressure
Represent these as nearly as possible by means of an

empirical formula.
9.
Amagat
experiments on
gave the following results
air at
very high pressures
EMPIRICAL FORMULAS
119
Represent these by an empirical formula.

10.
The
current through the field coils of a certain
dynamo was
varied and the voltage generated
by the
machine simultaneously measured, as follows:

11.
The
specific
gravity of dilute sulphuric acid at
different concentrations
is
given in the following table
120
pyknometer being tested for evaporation was

allowed to stand in a desiccator and weighed at intervals,
12.
as follows
Sept. 30
15 P.M.
44.4226 grams
4 00 P.M.
.4223 grams
11: 00 A.M.
.3855 grams
3 30 P.M.
.3821 grains
8 00 A.M.
.3695 grams
4 00 P.M.
.3622 grams
Oct. 2
Oct. 3
Find the most probable weight at noon October

13.
1.
Simultaneous observations were made upon two
connected variables x and y with the following results

14.
Following are vapor pressures, in
mm. of mercury, of
methyl alcohol at various temperatures
EMPIRICAL FORMULAS
121

16.
Assuming the form

Ar
n = logN
!
log
in
which n
is
per cent, and s
the data of Ex. 10, Art. 30.

16.
The
104-5
log
15
is
2
a log ,104-3
,
15
grade, deduce
N and a from
Plot the curve.
following average heights
and weights
for
men
35 to 40 years of age were compiled by the medical director

Mutual Life Insurance Co.
of the Connecticut
122
17.
The
Society for the Promotion of Engineering
cation reports
Try
its
growth
to calculate the
in
membership as follows
most probable membership
Edu
in
1915
from these data.
Try to represent the data plotted

by means of an empirical formula.
18.
19.
The
head
Ex.
8, Art. 30,
measurements give the
following
length of the
in
average
in schoolboys at different ages (West,
Science, Vol. 21, 1893)
EMPIRICAL FORMULAS
123
Records of the magnetic declination (departure of

compass from the true north) at 25 N. lat., 110 W. long,
20.
over a series of years are as follows (U. S.

1905)
Mag. Tables
for
CHAPTER
VII
47.
Relative Reliability of Observations.
Weights.
We
have hitherto regarded each one of a set of several

observations as having been made with equal mechanical
refinement, care and skill, and the results as meriting,
same degree of confidence. This assumption

however, far from the truth. The position of a
therefore, the
is
often,
example, as measured with an engineer
star, for
is
less reliable
meridian
than
circle;
observations
would be
made by
if
a tired research worker in a cold,
drafty laboratory are not worth as

series
made by
s transit,
measured with a large

and the results of a series of difficult
it
much
when
Again, the mean
the same person
favorable cpnditions.
careful observations
upon a quantity
is
as a similar
rested
and under
of a long series of
certainly of
more
value than the result of a single measurement upon the
same quantity.
It
is
therefore evident that, in practical work,
necessary to
reliability
by
employ some means whereby
may
be taken into account.
using a method of adjustment in
it
is
differences in
This can be done
which the more
trustworthy results are allowed to have more influence
upon the
final
most probable values than the

124
less reliable
ones, thus giving each result a degree of
125
prominence pro
portional to its reliability.
To
accomplish
this, it is
the practice of observers to
assign to different observations, numbers,
which are sup
posed to represent their relative degrees of reliability,

and which are called weights.
Thus an observation to
which the weight 3 has been assigned

merit only half as
with the weight 6
much
considered to
is
attention in the adjustment as one
etc.
In order to have some basis of estimation, we
an observation
the
mean
of a certain
number
of observations considered
as having standard or unit weight,
and
number is
The assign
this
the weight of the observation in question.
ment
may regard
of given reliability as being equivalent to
an observation means that
of the weight 10 to
the opinion of the observer the result
is
the average of ten observations of unit weight.
standard of trustworthiness
it
may
it
Any
be taken as a unit, but
should be such as to render the weights of
servations referred to
in
as trustworthy as
all
the ob
simple, whole numbers.
It
is
to
be remembered that weights are purely relative quantities.
The assignment
of weights to the several observations
is a task demanding the exercise of skill and

If each observation is actually the
careful judgment.
of a set
mean
of several elementary observations
the same kind, the matter

in this case a
there
is
wise,
and
especially
is
and
all
comparatively simple, since
numerical basis of estimate.
when
kinds, the assignment
is
are of
Other
the observations are of different
not so easy.
The problem
pre-
126
sents
many
analogies to that of giving numerical grades
to pupils.
Like other processes of the

servations cannot be covered
It
may
sort,
the weighting of ob
by any
set of definite rules.
be suggested that the observer should note and
record in detail the peculiar circumstances,
if any, attend
each
observation
or
of
set
observations which is to
ing
enter into the final adjustment, and allow no source of

unusual disturbance to go unnoticed. Often it is well to
assign weights at the time of the observation, while all
the circumstances are fresh in the mind, but this should
not take the place of recording the circumstances. It

sometimes happens that some one else examines the original
notes and prefers to assign weights for himself.
I recall
a case of this sort, in which the weighting depended solely

upon the records which the observer had kept of the
weather conditions prevailing at the time of each experi

ment. This was because wind and fluctuations of tem
perature were causes of marked disturbance in this par
ticular work.
48.
Adjustment
of Observations of
Unequal Weight.
In adjusting a set of observations to which different weights
have been assigned, we have but to remember that the
weight
signifies
equivalent in
It
is
that the observation in question
importance of
is
the
observations of unit weight.
therefore necessary only to repeat the corresponding
observation equation
times,
and then proceed as usual

That is, if the
with the reduction to normal equations.
first
observation has weight
etc.,
then simply write the
2,
the second
127
5,
the third 3,
observation equation twice,
first
The
sum of
the second five times, the third three times, etc.
number
of observation equations
now Sw,
is
the
the weights.
simple illustration of this
is
servations on a single quantity
q.
sn
with weights w\,
2,
wn
the case of n direct ob

If
the results are
Si, s%,
the most probable value
as deduced on the above principle
is
w ns
\~W n
mw = *^
(62)
2w
This
is
equal,
called the weighted
it
mean.
the weights are
becomes simply the mean.
With observation equations

several
If all
of the first degree involving
unknowns, the process can be effected by
multiplying the expression for each residual by the

cient of the
unknown contained
first
coeffi
therein (as in the rule at
the close of Art. 34), then multiplying by the weight of
the corresponding observation, adding the results and
equating the
In this
sum
way each
to zero, to form the normal equation.

residual
is
represented in each normal
The
equation a number of times equal to
its
same thing may be attained by
multiplying each
first
of the original observation equations
of its weight
weight.
by the square root

and then proceeding with the reduction
128
These square roots need only be indicated, by

means of radical signs, as they will disappear on re
as usual.
(Let the student show
duction.
why
the square roots of
the weights should be thus used, and not the weights

themselves.)
In the reduction of observations upon quantities limited
by
conditions (Art. 40)
it is
evident that the equations of
condition are not to be weighted, but only the observation
In the process of adjustment, the weighting

should be introduced after the conditions have been
equations.
involved in the observation equations, but before the re
Some
duction of the latter to normal equations.

following examples
of the
will illustrate this.
EXERCISES
49.
1.
of a line
Measurements were made upon the segments
AB, formed by
Mean
Mean
Mean
Mean
Mean
Mean
points C,
upon
it,
as follows
of 2 observations
on
AC =
45.10
of 3 observations
on
AD =
77.96
ft.
of 2 observations
on
CD =
CB =
32.95
ft.
of 3 observations
of 2 observations
of
4 observations
on
DB =
on AB =
on
ft.
98.36
ft.
65.55
ft.
143.55
ft.
Find the most probable values of AC, CD, DB.

2.
In one time-observation with a transit instrument,
only five of the nineteen lines of the reticle were used,
Nos.
2,
5,
10,
15,
18.
viz.,
second observation employed
all
What
the lines.
129
can be said as to the relative weights
two observations, the method

the same in both cases ? (Fig. 9.)
of the
of observing being
In determining the constants of a balance, it was

borne in mind that the instrument was to be used re
3.
peatedly for the weighing of an ob

ject varying
in the
always
measured
load
twenty-five
45
2.402
of
weight but
neighborhood of 43 to
Hence the
45 grams.
of
in
slightly
grams,
scale
was
sensibility
with
times
giving
divisions
mean
FIG. 9
milli
per
gram, and only four times with zero load, giving a
mean
of
probable
Ex.
2.767
scale
values
of
Determine
divisions.
the
balance
constants
most
the
(Art.
35,
2).
4.
Draw
tractor,
a triangle and measure
its
angles with a pro
one angle being measured but once, the second
three times, the third eight times (or
some other
set of
unequal numbers), all the measurements being made

Introduce the necessary condition, assign
differentially.
the proper weights and deduce the most probable values of
the angles.
6.
The
tions
in
following pointings were
the
triangulation
of
cm. direction theodolite (U.

1904)
made
California,
S.
at three sta
using
a 50-
Coast Survey Report,
130
Adjust for the most probable angles.

6.
The range
of
magnitude
of the variable spectroscopic
binary star a Geminorum was measured by a selenium
photometer on different nights as

Astrophysical Journal)
(Stebbins,
Find the weighted mean

7.
follows
of these observations.
Following are results from precise leveling in Texas
(U. S. Coast Survey Report, 1911).
The
weights assigned
are inversely proportional to the squares of the distances
between the
stations.
131
Adjust for the most probable elevations above the lowest

station in the
list.
Experiments were made for the purpose of rating a

Price current meter, used in measuring the velocity of
streams. The data are the velocity V of the current in
8.
feet per
of the
second and the number
R of revolutions per second
meter (Raymond, Plane Surveying).
Assume a
linear relation
and deduce the two constants.
132
9.
line.

Four points, A,B,C, D,
lie
consecutively in a straight
The following distances are measured with a steel tape.
Apply the principle that, in chaining, the weights of

similarly measured lines are inversely proportional to the
squares of their measured lengths, and adjust the above
values accordingly.
10.
Zenith telescope observations were
Station, Virginia,
upon the
made
at Roslyn
latitude of that station with
various pairs of stars, as follows (Chauvenet, Practical
Astronomy).
The weights were
assigned from the
and the precision with which the

employed had been measured.
of observations involved
declinations of the stars
Find the most probable
number
latitude.
133
(Adapted from Chauvenet, Practical Astronomy.)

of the U. S. Coast Survey, angles were read
At a station
11.
on each of four other
stations,
A, B, C, D, as follows
Adjust for the most probable angles.

12.
Spectrographic radial velocity measurements were
made upon
lines
on
the Orion nebula, using different spectrum
different
Bulletin No. 19)
dates,
as follows
(Lick Observatory
Find the most probable radial velocity.

13.
certain
critical
coefficient
measured several times with
of
expansion
different apparatus.
Find the most probable value from these data.
was
134
14.

The
following data are right ascension corrections
to the Berlin Jahrbuch
made by
the photographic transit
at Georgetown Observatory for the star f Ophiuchi on

different dates.
Find the weighted mean.

(Adapted from Wright
Adjustment of Observa
following trigonometric levelings were made
between two terminal stations A and B, as follows
15.
tions.)
The
By
precise spirit leveling,
above B, which
may
was found to be 39.05 meters
be taken as correct.
heights of the other stations above
50.
results occur
tempted to
is
is
It
sometimes
in the course of a series of measurements,

which are so doubtful that the observer is
reject
their weight
Adjust the
accordingly.
Wild or Doubtful Observations.
happens that,
135
them
In technical language,
altogether.
so small as to be seemingly negligible,
a question whether their retention
harm than good.

The doubt may
from the existence
arise
disturbing conditions,
may
known
and
it
not do more
of unusual or
to the observer.
On
one
occasion I was making a quantitative analysis to determine
the exact concentration of a solution, and during the proc

ess of drying, accidentally spilled a
My
water into the residue.
few drops of hydrant

was to be an
final result
average from the analyses of several specimens, and the

accident would unquestionably vitiate the result of this
observation
culty
and
but the specimens were obtained with
could
ill
afford to spare
the result to be rejected or not
Again, suspicion
between the
may
any
of the data.
diffi
Was
be due to a marked difference
result in question
and
all
the others of the
This does not refer to mistakes (Art. 9), which may

usually be easily rectified. To the observer s best knowl
set.
edge, the doubtful observation deserves as
as the others, having been
made with
but he dislikes to retain
it is
it,
as
much weight
the same care;
so far out of agreement.
136
The former
class of doubtful observations should, in
the opinion of the writer, be rejected unless some idea of

the extent of the disturbance can be obtained and due
made
correction
cited
was to
for
it if
test the
What
necessary.
I did in the case
hydrant water and ascertain that
the
amount
be
sufficient to affect the result at all seriously;
few drops would not

but I
of solids contained in a
gave only half as much weight to this observation as to

the others.
With the
latter class the case is
more doubtful.
because a result differs from the others

it is
any farther from the
ber of observations
is
is
truth, especially
small.
Just
no proof that
when the num
In casting out such a
result,
be throwing away his most valuable observation.

Certain criteria have been proposed for deciding whether
wild
to retain or reject a
observation, based upon the
one
may
"
"
law of error distribution.
Probably the best decision

be based upon the observer s judgment, it being borne
in mind that results of observations should not be tampered
will
Where wide
with unthinkingly.
be
if
well,
sufficient
it
will
possible, to continue the observations until a
number
distribution with
51.
deviations occur,
are accumulated to
show the law
of
some distinctness and symmetry.
The Precision Index
h.
It
was pointed out
in
Art. 28 that the quantity h in the error equation has to
do with the precision
of the observations (Art. 13),
the greater the value of

cated,
may
h,
the greater
thus be termed the
is
"
and that
the precision indi

precision index
"
"
or
measure of
to inquire
precision."
what connection
137
We
are here naturally led
exists
between the precision

For, if we have
index and the weight of an observation.
two
measurements, one of which
sets of
is
more
precise
than the other, the value of h belonging to the error dis

tribution in one set will be larger than that belonging to
the other
while at the same time the weight of one ob
servation from the
from the second

Let
hi
and
Ci
set is greater
first
than that of one
set.
be the constants in the equation of error

and let Wi be
distribution corresponding to the first set,
the weight of an observation from that
them
all
and
to have equal weight;
set,
let h%, c2 ,
corresponding quantities relating to the second

probability of
an error x occurring
yi
Cl e-
supposing
w2
be the
set.
The
in the first set is
(63)
Let the value of the precision index corresponding to a set

in which the observations are of unit weight be h.
This
may
"
be called a
standard
index,"
value can as yet be assigned to

the
first
set is
error x
set, in
is
y
Therefore the probability
first set is
which
is
that of
its
though no absolute
observation from
An
equivalent in worth to Wi observations
from the standard
an
it.
giving
each of which the probability of
Wa ,
&*"*.
y\ of
the error x occurring in the
occurring Wi times in the standard set,
138
The
error x being supposed the
these equations holding for

gives at once
same
all
in (63)
and
(64),
and
values of x, comparison
hi
Likewise
referring to the observations of the second set, having
weight
wz
That
2
hi
is,
2
:
h%
h$
2
:
= wi
w% w%
:
(65)
or the weights of observations are in proportion
to the
squares
of their precision indices.
In order to illustrate this principle, let the error distri

first set be represented by A, Fig. 10, and
bution of the
FIG. 10
that of the second by B.

of
measurements.
a and b from the
From
A represents the more precise set
Let the points of inflection be distant

?/-axis in
the two curves respectively.
(27), Art. 28,

1
or
b
Then from
a.
(66)
(65),
(67)
139
thus possible, by means of a study of residual

curves, to estimate the relative weights of observations
It is
made by
or
by
with different instruments,
different processes, or
different observers.
In the next chapter (Art. 59)
will be presented a mathematical means of obtaining the
same information, without plotting the curves.
This does
not, of course, refer to the weighting of individual observa
tions of the
ment
same
set,
which must depend upon the judg
of the observer as to the conditions existing at the
time.
52.
Statement of
General
The
Squares.
ciated in three
the
Principle
Least
of
principle of least squares, already
ways adapted
to increasingly complicated
cases of adjustment (Arts. 29, 31, 33),

in its general form,
enun
which includes
all
may now be deduced

the others as special
cases.
Let a
series of
n observations be made, whose weights
are respectively w\,

Pi)
pz>
")
Pn-
w<z,
The
w n and
,
the residuals be
^~w
which
and h are the precision constants corresponding
to an observation of unit weight.
occurrence of
Y=
let
probabilities of these residuals are
(64)
in
2/12/2
all of this
yn
wl
The
probability of the
particular set of residuals
+ W2+
+Wn e- h2(wipl*+ wZ +
is
+ w npn
(68)
140
The most probable set of residuals, and hence those

determined by the most probable values of the unknown
quantities involved in the observations, are those for which
maximum, and hence those for which 2(wp ) is a

minimum.
The most probable
The general statement follows
values of unknown quantities connected by observation equa
is
tions to
which weights have been assigned are those ivhich
sum of the weighted

a minimum. The meaning of
will render the
"
squares
The
is
squares of the residuals
the
term
"
weighted
obvious from the above.
rules of Art. 48 for the
adjustment of weighted ob
servations might have been deduced from the principle
as above stated, in the same
was made
33, 34).
manner
as the deduction
for the simpler case of equal precision (Arts.
CHAPTER
VIII
PRECISION AND THE PROBABLE ERROR

53.
of
Discontinuity
There
the Error Variable.
is
one point in the foregoing discussions of the law of error

In all of the mathematical
that has not been emphasized.
work, we have treated the error as
if
it
were a true con
tinuous variable x, which might have any value whatever
from
oo to
oo
But
to assume this would be to assume
an
infinitely
To
illustrate the fact, let
quantity
is
minute graduation
an angle.
of our
measuring scale.
us suppose that the measured
If the error were a continuous
variable, successive measured values of the angle need
not differ by so much as a billionth of a second, yet might

be different; and the probability of any particular error
out of the infinity of possible ones would be infinitesimally
small.
It is thus seen that the variable error x, instead of
varying by infinitesimal increments dx, really has equal

discontinuities A, which represent the smallest
finite
fraction of a unit in
pressed.
which the measured
On a surveyor
s transit,
results are ex
for example,
is
usually
one minute for single angle-readings; while with the

micrometers used on large equatorial telescopes, angular
measurements are made which

dredths of a second.
141
may
be expressed in hun-
142
The
may then be represented as a sort of

with
stairway
equal treads and unequal risers, and the
errors considered as falling into compartments correspond
error curve
/
/
the
ing
to
A s,
just
several
as do the
shots in the
|\
experiment
IA
1L
The width
target
of
Art.
of one
of these error
"
com
being
height y,
it
"
partments
FIG. 11
and
its
may be looked upon as a narrow strip of finite area i/A.

y is the probability that an error will fall into the com
partment to which the ordinate y corresponds. Let us
imagine all the strips placed end to end, the total length
being
*Ly
1,
since this
is
the
of all possible errors (Art. 16).

strip
is
the area of the curve
A-e
sum
of the probabilities
But the area
of this long
Ve.A
2;= +00
- h2 *2
z=0
and
=4
A
2V.
being small, the summation 2 in (69)
(69)
is
purposes represented by the definite integral
for practical
PRECISION
I
= f
ô
Whence, from
The
integral
(69),
is
a function of
h,
and when we have de
h,
we
between the two constants
Value of the Integral
54.
The
h.
preceding
(70)
(71)
shall
/,
have found the
and h
which was referred to in Art.
tion,
and
w *(fo.
termined what function of

lation
143
of the error
re
equa
28.
and the Relation between
evaluation of the definite integral I of the
article is
worked out
in
Note C
of the
Appendix,
this being a
problem belonging to the theory of definite

and
an interesting example of a method often
integrals
employed in such cases. The result is
|l.
(72)
The student who does not
care to follow out the proof may

the
result
verify
by plotting the function for two or three
chosen values of h and integrating the curve with a planimeter. However, the note referred to is not difficult to
read,
and students are advised to do
so.
Substituting the value of I here obtained in (71), the

expression for c
is
found to be
-,
VTT
and
is
therefore proportional to h.
(73)
144
We may now write the error equation in a final and more

satisfactory
form
in
which the law
upon the
-*>-"
(74)
of error distribution is
which
scale-interval A,
is
made
to depend
readily obtained from
the recorded observations, and upon a constant h which
we have come to refer to as the precision index

which, as will be seen
later,
(Art. 51)
and
can also be calculated from the
results of the observations.
55.
The
Probability of
an Error Lying between Given Limits.
Probability Integral.
theory of errors
will lie
is
An important problem
between two given limits Xi and
obtained in terms of the precision index
X
h.
This
z.
may be
For, the result
is merely the sum of the probabilities

between Xi and
z which is, by (74)
sought
in the
to find the probability that an error
of all errors
A by dx as in equations (69), (70), in order to

convert the summation approximately into an integral,
or replacing
XX2 e
-vx*
dx-
7,
(*Xl
I
-v/~^0
e- ht * dx.
(75)
PRECISION
The
is
145
definite integral occurring in (75), viz.,
2
by a substitution. Let kx = z, h?x =
when x = X z = h X. We then have
simplified
hdx
dz
2
,
fhX
fUX -
(76)
This expression
limit
upper
an error
the
to
less
is
hX.
will
useful
is,
is
It
lie
question
bility that an
that
and
integral,
ability
evidently
function
the
expresses
and
between
of
probability
+ X.
precision, the value
of
the
called
commonly
is
prob
the
that
As applied
of
itself
than that of 2F, which is the proba

error will lie between +
and
X,
that a measured result will be within
of
the true value.
Our problem now
Y from
requires that
the given value of h X.
expressed directly as a function of
uated through the use of

cal
work
student
is
is
we be
able to calculate
cannot, however, be
hX, but must be
infinite series.
eval
This mathemati
D of the Appendix, to which the

Tables of the values of the integral,
given in Note
referred.
thus calculated, are standard, and given

the theoryôf errors.
values of( 2
ment hX.
re
in every book on
In the accompanying table the
given,
corresponding to the
argu
146
56.
-
//
Calculation of the Precision Index from the Re-
siduals.
The
table in the preceding article enables us to
find the value of 2
Y corresponding to any given
value of X,
providing the precision index h is known. That is, if we

have h, we can find from the table the probability that an
error will not exceed a given value
Reciprocally,
if
X in numerical value.
the value of 2
corresponding to a
PRECISION
given limiting error
the same
as a
from
can be determined by any means,

and hence of h, just
table will give the value of h
number can be found from

its
147
tangent,
by
its
logarithm, or an angle
interpolation.
This
plished from a study of the residuals
observations
is
if
may
the
be accom
number
of
large enough.
For, the probability 2
that a residual will
lie
between
+ X and
may be obtained by finding what propor

tion of them do lie between these limits.
By choosing
X, as many values for h X may
be found, which may be combined like observation equa
several different values of
tions for the
While
unknown quantity
h.
this is rather too laborious a
purposes,
clearly in
it will
mind the
measurements.
method
for practical
be found a very useful means of getting
We
relation of h to the precision of the

shall therefore
apply it, by way of

illustration, to the following results of 144 measurements
upon the length of a line. Of these:
RESIDUALS
RESIDUALS
148
The
interval
An
examination of the residuals gives the following
facts
between successive residuals
is 0.1
foot.
The numbers
in the
column headed 2
values of the fraction
55
^V
;
144
values of
hX
e.g.,
144
are simply the
0.3819, etc.
The
corresponding to these values of 2 7, ob
tained from the probability integral table, are as follows
any one
By
of
which
will give
using the above values,
tion equations for
h,
thus
an approximate value
we may
of h.
write simple observa
PRECISION
0.1 h
0.2 h
=
=
149
0.353,
0.616,
etc.,
which when reduced in the usual manner (Art. 34) give

as the most probable value
h
2.33.
Using this as the value of h and 0.1 as that of A, and sub

stituting
them
in (74), gives as the equation of error distri
bution for this case
0.1314 e~ 5A
Let the student plot this curve and the actual distribu
tion of residuals together on the
of comparison.
five or ten
The
ordinates
same sheet
for the
had better be
purpose
laid off
on
times as great a scale as the abscissas, for con
venience.
Approximate Formulas for the Precision Index.

The foregoing is doubtless as accurate a method of ob
57.
taining the precision index h as could be desired where the
number
of observations is large,
and where
it
can therefore
be assumed that the residuals distribute themselves in

accordance with the error law.
for practical purposes,
The
It is
however too laborious
and can be replaced by shorter
one here presented depends, in fact,

upon the same principle as that used in the foregoing cal
methods.
first
culation, being simply
more
direct.
Let there be n observations made upon the unknown
150
quantity or upon functions of it, the errors being xi,

x2
xn
From (74), the probabilities of these errors
,
are, respectively,
= -=e~ h \
fli-^
yi
v.-i-a
and the probability
of the given set of errors is
(77)

these observations
is
of h that can be afforded
the one giving
probable distribution of errors,
equivalent to the statement that
which
condition
is
to be a
by
the most
rise to
is
maximum.
Hence, regarding h as a variable and obtaining the condi

tion for Y a maximum by differentiation of (77), we have
h=
whence
This would be adequate
and does very

servations,
true errors.
if
well in
any
(78)
if
the true errors x were known,
case,
where there are
many ob
we simply use the residuals p instead of the

The discrepancy between (78) and the value
PRECISION
151
obtained by using the residuals is discussed by

writers at some length, and it is a question whether
many
it
de
serves such attention, inasmuch as only an approximate
value of h
So:
is
is
is
S/o
may
be easily seen that
greater than
squares
It
usually required.
Z/>
to be a
since by the principle

minimum. Hence if Sp 2
,
to be
something must be done to
used instead of Sx in
of least
is
(78),
reduce the numerator as well as the denominator.

general practice
is
make
to
it
instead of n, a pro
The formula
cedure which has some theoretical support.

for h
now becomes
n
This formula
is
what
use
somewhat
is
squaring
all
is
(79)
of the greatest
known
tion of
importance in the calcula
as the probable error (Art. 58).
owing to the
laborious,
the residuals.
The
Its
necessity of
Another formula
for h, first
used by Peters, can be derived upon the following reason

ing.
is n.
Then
The total number of errors, both + and
nx be the number of errors having the particular value x,
,
if
their probability is
y=^n = ~eVTT
Let us consider only

is
the same as that of
out signs.
The sum
+ errors,
all
h* x
(80)
the average value of which
the errors,
of all the
and
errors
is,
taken with
from
(80),
152
or approximately (see Art. 53),
The average
of the
(disregarding sign),
and hence
errors,
-f-
of all the errors
therefore
is
(nx x)
n
whence
\n
h
This formula,
h^TT
(81)
like (78), is in
terms of the errors
x.
In
order to reduce (78) to the expression (79) for h in terms of
the residuals
If
p,
the numerator was reduced in the ratio
we apply the same
placing the x
which
process to (81), at the same time re
by the p
s,
we obtain as the analogue of
be referred to as Peters formula for

be called the standard formula for h.
will
(79) will
It is to
h,
(79),
whereas
be noted that the above reasoning applies only
to observations of equal weight.
The
question of weight,
as related to precision, will be introduced in Art. 59.
We
have
heretofore treated the precision of observations in a
more
58.
The Probable Error
of
an Observation.
PRECISION
or less abstract and relative way,
153
and the need
In short,
ap
more concrete and tangible expression for it.
we desire something that will convey to the
mind an
idea of the accuracy attained, in terms of the
units of
measurement used.
is
parent for a
This has been secured in
several ways.
One
of the simplest quantities of this sort is the average
residual,
taken without reference to
to the precision index
is
sign.
Its relation
obtainable from (82), which
gives as the average residual
(83)
or
if
is
very large, approximately,
n
which
is
equivalent to (81).
Again, there
residual,
is
the virtual or radical
which from
mean square (R.M.S.)
(79) is
(85)
n
or
if
is large,
approximately,
The
that
concrete significance of this quantity

it
(86)
lies
in the fact
represents the abscissa of the point of inflection
on the error curve
(Art. 28,
Eq. 27).
154
The most approved expression used for this purpose is,

however, the probable error. In Art. 55 it is shown how to
calculate the probability that
an error
will
not exceed a
X, providing the precision index h
given limit
known,
is
the table of the probability integral being used in the cal

culation.
We may now
Designated by
give the following definition.
the probable error of
e,
an
observation is
such that the probability that the given observation differs
from
the truth
by an amount numerically
equal to the probability that

cally greater than
More
briefly,
differs
error
as
it is
just as likely to be less than
is
lies
In the long run, half the errors

exceed
by an amount numeri
to be greater
the probability that an error
will
or in other words,
between
will lie
and
within
e is
and
e,
J.
half
it.
Therefore,
c is
that value of the limit X, appearing in
the argument h X, which corresponds to 2
Y =
in the table of the probability integral.

in this table gives as the value of the
which 2
is
e.
any
the probable error
it
than
less
F =
0.5000
Interpolation
argument
hX
for
0.5000,
he
whence the probable error

6
0.4769,
is
_Q^6.
h
(87)
Using the standard formula (79) for
0.6745
-,
h,
this gives
(88)
PRECISION
in terms of the
sum
of the squares of the residuals;
we obtain
using Peters formula (82),

the probable error,
sum
in terms of the
Peters formula
155
Peters formula for
(89)
of the residuals without sign.
it will
be seen that when n
proximately,
e
is
From
large,
= 0.85^;
to
equal
85 per
observation is approximately
cent, of the average residual,
This simple rule
sign.
an
is
ap
(90)
or the probable error of
or
taken without
sufficiently accurate for most
practical purposes, in the case of a long series of observa
tions of equal weight.
The notation by which probable

uses the double sign.
errors are expressed
For example,
if
the mass of an
obtained by weighing, is stated as 24.830726

0.000014 grams, this means that the probable error of
the weighing is 0.000014 gram. This quantity would be
object,
by taking a series of weigh

on the same balance under the same conditions,
obtained, as explained above,

ings
finding the residuals,
59.
When
and applying
(88), (89) or (90).
Relation between Probable Error and Weight.

the several observations of the same series are
assigned different weights, the probable error of a single
observation has no significance without further qualifica

tion, since
the precision index
h,
and hence the probable
156
error, is
supposed different for the different observations.
We may
express both, however, with reference to obser
vations of some selected precision, as those which have
We
been assigned unit weight.

precision index and by
and
of unit weight,
by h the
shall designate
the probable error of observations
refer to
them
also as observations of
Other observations whose weights

have precision indices hi, fc,
and
,
standard precision.
are
w\ t
wz
probable errors
We have
(Art. 51,
(87),
2,
already deduced the relation between h and
h%
hz
for fe
^3
It is
Eq. 65).
hi
From
ei,
= V Wl ^w2 ^Wi
:
:=-:-:-:
l
63
Combining these proportions,

ei:
e3
=-7=
*Wi
~j==
^w2
-7=
^wz
~,
(91)
which expresses the very important principle that the

probable errors of different observations in the same series
are inversely proportional to the square roots of their weights;
or reciprocally, the weights of observations of the same kind

are inversely proportional to the squares of their probable
errors.
To
if the probable error of a measured

obtained
by one method is found to be only
quantity
one-half as great as that obtained by a less precise method,
illustrate this,
then the weight assigned to the former in combining them

should be four times that assigned to the latter. In other
PRECISION
157
words, one observation by the former method

four
made by
is
worth
the latter.
be the probable error of an observation of unit

weight, found from (88), (89) or (90), then by the foregoing
If e
principle, the probable error of

is
given by
,
This
will shortly
an observation
of weight
(92)
-f:.
VHJ
be seen to have an important applica
tion to the finding of probable errors of adjusted or
probable values of
If
is
unknown
most
quantities.
the probable error of an observation of unit weight
to be calculated from a series of weighted observations,
we may generalize the reasoning of Art. 57 as follows.

The weights are w\, wz
and the corresponding precision
,
indices
hi, Jh,
h being the standard index,
*.The
probabilities of the respective errors are
now
given by
158
The
joint probability, corresponding to (77),
is
now
&
Y=
from which, by the same reasoning as that leading to
and
(79), the
standard index of precision
Instead of (88)
we now have, by
(78)
is
substitution of this
new
expression for h in (87),
= 0.6745 J^2-,
\ n
(94)
the important standard formula for the probable error of
an observation
of unit weight, as obtained
of weighted observations.
ming
plied
from a
In this formula, before sum
the squares of the residuals, each square
by the corresponding weight;
residual
is
series
is
multi
or, otherwise,
each
multiplied by the square root of the corre

(See Art. 52.)
sponding weight.
The same
modification
formula (89) to adapt
it
may
be
made
in the Peters
to weighted observations, giving
(95)
(n-l)
or
if 7i is
large, approximately,
which corresponds to
.85^^,
n
(90).
(96)
PRECISION
159
EXERCISES
60.
7701
1.
a,
Two
was
of a special
one of which, No.
specific gravity bottles,
of the ordinary type,
and the
other,
improved design, were each
five times at the
filled
No. 7701
c,
with water
same temperature, the following being

made on the same
the results of the weighings, which were
balance in the same manner
No. 7701 a
No. 7701
42.602818
45.345518
42.604108
45.345852
42.603512
45.345597
42.602062
45.346437
42.602947
45.346219
Find the probable error of a single filling and weighing

with each of the two bottles, and the relative weights of
a single observation in the two cases.
2.
Eighteen measures of a horizontal angle were made

of a large Coast Survey theodolite, as follows,
by means
the observations being of equal weight

13
31
17".6
13
31
20".4
21
.5
20
.9
19
.0
23
.5
21
.5
18
.4
26
.2
14
.2
17
.1
21
.0
22
.1
21
.8
20
.1
22
.4
17
.9
17
.6
160
Find the probable error of a single observation of this

series by means of each of the formulas (88), (89) and (90).
Regarding the mean as an observation of weight
find the probable error of the mean.
3.
Find the probable error
experiment of Art. 11, Ex.
of
one shot in your own target
1.
Find the probable error of one observation

series of measurements which you made upon a
4.
Art. 11, Ex. 2.

6.
18,
Also, find the probable error of the
Six separate researches,
by
in the
line in
mean.
different observers,
upon
mean
with
the velocity of light gave the following
their probable errors, in kilometers per second
298000
1000
298500
1000
299930
100
299990
200
300100
1000
299944
50
results,
:
Assign the proper relative weights and find the probable

error of an observation of unit weight.
Also, regarding the weighted
of weight
Explain
is
mean
as an observation
Sw, find its probable error.
why
the answer to the
not 1000, supposing the
unit weight.
less precise
Give reason
From
first
first
part of the problem
observation to be assigned
the answer to the second part, do the
observations add to the value of the whole?

for
your conclusion.
PRECISION
6.
161
The constant of a Babinet compensator is determined
by measuring the
two successive dark
distance between
Micrometer readings
bands as seen through the analyzer.

were taken as follows
:

ence in readings;
7.
of
of the
one measurement of the
Ten measurements were made upon
the magnitude
of a certain bright star, with the following results
0.600
0.470
.460
.483
.477
.475
.500
.490
.467
.475

mean.
8.
of
one measurement and of the
Syntheses of carbonic acid gas
kinds of carbon by
differ
mean.
Dumas and
made from
different
Stas gave the following
162
(Freund, Chemical Composition). The numbers

represent the percentage of carbon in the gas.
results
Find the probable error of one determination and
of the
mean.
9.
both
In a
series of
steel
and invar
were found (U.
S.
base line measurements
made with
tapes, the following probable errors
Coast Survey Report, 1907)
Averaging these, find the relative weights of base

measurements made with these two tapes.
10.
Apply Peters formula (95) to

an observation of unit weight
error of
14, Art. 49.
line
find the probable

for the data of
Ex.
PRECISION
Probable Errors of Functions of Observed Quanti
61.
An
ties.
sion
is
"
error
able
163
important phase of the subject of preci

what may be termed the "propagation of
and
illustrated
the
of
error
measurement,
is
by an
diameter
what
e;
example
a
of
circle,
The prob
obtained by
the probable error
is
Or
the area calculated therefrom?
of
generally, given the
probable error of a measured value of a quantity, to

find the corresponding probable error of any function of
that quantity.
Let the measured quantity be
q,
and the function
Let an observation be made upon q with error
x,
and
let
the corresponding error affecting the function Q, as a
be X.
result of this,
Then
if
x be small, we have ap-
X :x = dQ:dq,
proximately
X=A.
or
dq
It
may now readily be
seen that
able errors of the measured q
E=
and
if e
and
E are the prob
of Q, respectively,
^-e.
dq
then
(97)
This may, however, be shown as follows.

x n are a series of errors committed in
If Xi,
2
,
measurements upon
q,
and X\,
errors in Q, then as above,
2,
X n are the resulting
164
*-fk
dq
X
*
or squaring
II
-xdq
7
^n
and adding,
(98)
Now
from
since the
(87), substituting the
or
and the
the probable error of q
are true errors (not residuals),
is
and that
value of h given in (78),
=0.6745
J*
of Q,
0.6745
from which
0.6745 2
0.6745 2
The substitution of
these in (98) with subsequent reduction
gives (97).
That is,
quantity
the probable error of a function of a single
is
equal
to the derivative of the
measured
function times the
probable error of the measured quantity.
For example,
q
9.67
0.02
if
the measured radius of a circle be
cm., the
computed area
is
Q =
wq
PRECISION
293.7663 sq. cm., and
0.02
In general,
of
probable error
is
a function of several
Q -/fa,
if x\,
x2
-, qt
q2 ,
qi,
is
E =
resulting error of
(/)
-,?,).
ft,
(99)
x are the errors of the respective values

t
simultaneously substituted in (99), the
given approximately by
is
Let there be a number

each giving
s,
is
(ri)
(ioo)
of series of observations
rise to
(100), viz., Xi, Xi,
which
measured quan-
*-lji*+g*+-+3hthe q
irq
1.215 sq. cm.
titles:
Then
its
165
n.
obtained from the
an error
upon
as represented in
Then approximately,
X s upon
squaring (100) and
omitting the product terms of the expansion. This omis

sion is justified by the fact that there will be in the long
run as
many
+ products
as
and they
will
be distrib
uted approximately in accordance with the error law,

and will hence practically cancel each other; whereas,
the square terms are
By
+, and must be
retained.
the same reasoning as that employed in the simpler
case following (98),
of
all
which (97)
quantities
e i}
may
e2 ,
we now
readily obtain
be regarded as a special case. The

e are the probable errors of measured
t
166
values of
qi,
q2 ,
qt ,
and
the probable error of
sulting from substituting these values
re
in the function (99) .
As special cases of importance, we may take the following

If
(a)
Q =
Kiqi
+Kq +
2
E = V&V + #
then
If
(6)
Q=ffgiV-fc
+ K,q
V+
lt
+Kfa*.
(103)
r
,
Let the student deduce these

62.
results.
Probable Errors of Adjusted Values.
The
dis
cussions of the probable error heretofore have been con

fined to the results of single measurements.
The
values
finally taken as the most probable, for the unknown quan

tities, from a series of measurements may, however, be more
trustworthy than that of any single measurement, and the
manner
us,
by
of their calculation
from the observations enables
applying the laws developed in the preceding article,
to calculate the probable errors of these adjusted values,
regarding them as functions of the observations.

As the simplest illustration, we shall first take the case
of direct observations of equal weight
tity q.
The
upon a
observation equations are
=
=
= 9,
si,
s2 ,
single
quan
PRECISION
there being
167
observations.
ra
be given in the form
may
m = -Si-\--s
n
This
+-s n
2 -i-
a function of the form (a) of the preceding article,
is
and the probable error is given by (103). Each observa

tion s has the same probable error, designated by (88),
e
= 0.6745 J-^
\n
Then by (103) the probable

mean m is
error of the arithmetical
-, (105)
n(n
which
is
the formula ordinarily applied.
tained at once from (88) and (92)
metical
mean
of
n observations
It
be ob
may
by regarding the arith

of
unit weight as an
observation of weight n, as suggested in certain of the
problems of Art. 60.
By
tions
a similar course of reasoning,
upon a
single quantity
if
there are
having weights Wi,w2
assigned, the probable error of the weighted

e mw
The
is
*
0.6745.
mean
wn
(Art. 48)
^l
(n
(106)
l)2to
general case, in which there are
n observations
of
upon functions of unknowns q if q%,

somewhat more complicated.* We shall deal only
different weight
t,
n observa
See article by the author, Popular Astronomy, Vol.
p. 239.
XIX,
168
with the usual problem of first-degree observation equa

represented
tions,
by
(38), Art.
The
34.
residuals are
then given by (39), from which their numerical values

must first be calculated. The probable error of an ob
servation of unit weight
is
now
residuals in the usual manner, or
=
which
0.6745
commonly taken
is
which certainly
by another formula
J^*V-,
differs little
from
many
(94)
/
satisfactory,
when n
(94)
/.
as a special case of (107), in which
(107)
more
as being
should be, large compared with
In
calculated from these
may
is,
and
as
it
be regarded
1.
kinds of work, the probable error of an ob
is known to the observer through

his
with
instruments, and need not be
long experience
calculated with reference to each series adjusted. At
servation of unit weight
sn
0i,
we
suppose the probable errors of Si, s*,

in equations (38) to be known, and designate them by
any
rate,
02,
shall
0n-
errors of mi,
ci,
2,
It
2,
is
now
,
ly
required to find the probable
which
may
be designated by
cz .
In the process of adjustment of such a set of observa

tions as here referred to, there arises a set of I most prob
able or normal equations, which, from the
at
them
(Arts.
34 and 48),
may
more conveniently,
of arriving
be symbolized as follows
+
etc., or
mode
+ Rtfnj) = 0,
169
PRECISION
.imi
2(6ms)
A<2.mi
+ Bim2 +
+B m +
2
,--..,
(I0o)
we
If
of
represent
members
the
then these
the
first
EI,
E s may
z,
the probable errors
respective normal equations,
each be expressed in two ways.
place, since the first
equation
member
of the first
In
normal
is
2(aws)
its
by
of these
aiwisi
+ a w S2 +
2
4-
probable error EI, as a function of the
a nw ns n
given by
s, is
(109)
),
with similar relations for the other normal equations.

Again, the probable error EI of the second member of the
first
normal equation (108), as a function of the
given
is
by
(HO)
with similar relations for

(109)
s,
and
(110),
2,
E$,
and the other
E^
Then equating
we obtain
similar pairs,
the system
2(6
Ve
),
V= 2(rW).
2
6
These equations are of the first degree in ei 2 e 2 2
and may be readily solved for these values, the required
,
170
probable errors themselves being then obtained by ex

tracting the square roots.
The
actual process is often not as complicated as might

be supposed, especially when the number of unknowns is
R appearing in (111)
not large. The coefficients A, B,
are already known from the normal equations.
The
,
weights
are simple numbers, frequently
all
The
unity.
work
is greatly facilitated by the use of tables of squares

and square roots, and the slide rule. And it may finally
be remarked that, since no great precision is required in de
termining probable errors, superfluous decimal places
may
be dispensed with in the several stages of their calculation.
Another method of calculating the probable errors of

adjusted values from observation equations of the first
degree
63.
is
given, without proof, in the Appendix,
Note E.
Probable Errors of Conditioned Observations.
The laws developed
in Art. 61
make
possible the extension
of the foregoing processes to the case of observations
quantities limited
will
are
by
be remembered that the
first
upon
rigorous conditions (Art. 40).
It
ra equations of condition
used to obtain the value of m of the unknowns
in terms of the others, those values being substituted for
them
The
observation equations before adjustment.

probable errors of the quantities still involved in the
in the
observation equations
may now
in the preceding article.
errors of the
be found as explained
This being done, the probable

may be found as
eliminated quantities
functions of the others,
by means
of (102).
PRECISION
171
EXERCISES
64.
is
(The formula for double weighing on a balance
1.
W=p+
used,
left
ri
and
and
-,
r2
which p
is
right pans, respectively,
the values of
factor
sum
the
are the pointer readings
and
r\
of the weights
when
s is
ties,
on
object,
used was 0.0002753.
Find
of the
0.96
0.93
1.08
0.95
0.99
1.12
1.02
1.05
0.92
1.10
for the load
the probable error of one weighing, and of the
2.
is
2s
ten weighings.
object
the sensibility
same
For ten weighings

r2 were as follows
of the balance.)
The
in
mean
of the
(Does p need to be given for this purpose ?)
Given, the probable errors of two measured quanti

to find the probable error of their calculated
sum
or
difference.
3.
which are given

Art. 60, were made on the same balance and by
All of the weighings, the data for
in Ex.
1,
the same method as those giving

of this article.
The
latter
rise to
the data of Ex.
data refer to ten weighings of
which p = 17.423 g.
The former data, referring to bottle No. 7701 a, are for
fillings with pure water at 21 C., at which temperature
the
empty
bottle
No. 7701
a,
for
172
the specific volume of water
1.001957 cc. per gram.

Find the most probable capacity of the bottle at this
temperature, and the probable error of the result.
is
Find the probable error of a single determination of the

capacity, the above data being regarded as five deter
minations.
Given, the probable errors of the measured legs of a

right triangle, to find the probable error of the calculated
4.
hypothenuse.
Given, the probable error of a measured angle, to
6.
find the probable error of its sine, derived therefrom;

of its tangent.
Find the probable errors of the constants V and

calculated from the data of illustration 2, Art. 36 (p. 80).
6.
Find the probable errors
7.
of the constants a, b, c
calculated from the data of illustration
(From
8.
J.
1,
Art. 44 (p. 107).
The
P. Bartlett, Least Squares.)
weighted observations were
made upon
following
the differences of
longitude of four American observatories:
Cambridge
Washington 23 m. 41.041
42 m. 14.875
Cleveland
s.,
wt.
Cambridge
Columbus
s.,
wt.
s.,
wt.
s.,
wt.
Cambridge
Washington
Cleveland
The
47 m. 27.713
Columbus 23 m. 46.816
Columbus
5 m. 12.929
s.,
wt. 30
longitude of Washington being taken as 5 h. 8 m.
15.78
s.,
find the
most probable longitude
of each of the
other stations, with their respective probable errors.
PRECISION
9.
The probable
error of a single observation
angle with a surveyor

If
the angle
five
times and
six
known
s transit is
of a triangle
173
is
to be
upon an
1
4".
measured three times,
times with this instrument, calcu
late the probable errors of the
of
A, B, C, obtained by adjusting these measurements.

10.
Adapt Peters formulas
to the probable errors of
adjusted values, (105), (106); also to (107).

11.
If
the current in a galvanometer corresponding to
deflection 8
is
tan
6,
find the probable error of a
current determined from a deflection reading whose prob

able error
12.
The
is
following table occurs in a certain Coast Survey
report, referring to the probable errors of the various sec
tions of a base line, in millimeters.
174
Fill
out the last column and obtain the probable error of
the whole base
line.
The most probable
13.
13191.3417
length of the Stanton Base
value and probable error of

at least a seven-place table
The weights
14.
DAE,
are 2,
16.
measured angles, BAC, CAD,

Find the corresponding
respectively.
1, 5,
BAE obtained by adding the measure
The probable
.2,
and
(Omit unless
logarithm.
available.)
of three
weight of the angle

ments.
is
its
is
is
Find the most probable
0.0052 meters.
on a transit
error of a circle reading
of a pointing at a signal, O
What
.l.
is
the
probable error of a single differential angle measurement ?

16.
1,
The probable
and
on a mark being
find the probable error of an
error of a setting
of a circle reading,
e2 ,
angle measurement by the cumulative method, using n
turns of the
17.
eter
.l
is
;
circle.
The probable
0.07
and
mm.
of
error of a scale reading
on a cathetom-
of a setting of the telescope
an adjustment
probable error of the
mean
meters from the instrument.
of the level,
.07.
of ten readings
If
on a mark,
Find the
on a mark 2
you had to use a cathe-
tometer, would you analyze the probable error in this
way ?
How
would you find
it ?
Following are the results of three series of meas

urements on the combining weight of lithium, made by
18.
different chemists (Freund, Chemical Composition)
PRECISION
Diehl
....
Troost
Dittmar
Weight these
its
175
59.417
0.0060
59.456
0.0200
59.638
0.0173
and obtain the weighted mean and
results
probable error.
Five independent series of determinations of the

atomic weight of silver gave the following results (Freund,
19.
Chemical Composition}
107.9401
0.0058
107.9406
0.0049
107.9233
0.0140
107.9371
0.0045
107.9270
0.0090
Assign weights and obtain the weighted

probable
What would
be the
its
mean
a persistent error entering

Might the probable error of a
effect of
one of such a series?

weighted
mean and
error.
ever be greater than that of any one of the
observations entering into
it ?
Apply the appropriate Peters formula to finding the
20.
probable error of the weighted mean of the observations

of Ex. 14, Art. 49.
(See Ex. 10, Art. 60.)
The probable
21.
tions
is
found to be
error of the
0.1 per cent.
mean
of fifty observa
How many
servations would be necessary to reduce

cent. ?
it
more ob
to 0.01 per
APPENDIX
SUPPLEMENTARY NOTES
A.
Proof of the Necessary Functional Relation As

in Deriving the Error Law.
(Supplementary to
sumed
To deduce
Art. 27.)
the form of the function 0, such
that any set of values of
x2
x\,
...,x n
Z = Z!+z +
Let us add a small

,
new
will
it
quantity
to any one of the
from any other, say x s making

x/ = x r + e, x = x s
,
Values of these quantities
This
e.
sis,
finite
say x r and subtract
s,
(a)
simultaneously render
will
the
+o
that will render
will
not alter the condition
not alter the condition
3>
X=
0,
and hence
0, since, by the hypothe
This necessi-
these conditions are to be simultaneous.
tates that
[0(av
+ c) - 0(or +
whatever the values
r )]
of
Dividing through by
[4>(x
x r x a and
,
e,
this
e)
c.
may
be written
177
- 0(x )] = 0,
APPENDIX
178
Allowing
to approach zero, this becomes at the limit
or since x r
and x a represent any two
It follows at
once that, since the x
any manner among themselves, only

4>(x)
is
ax
a constant, say K.
4>(x)
That
tution in
the
first
may
s,
in general,
be varied in
so condition (a) holds,
and
c.
(6) jointly,
since substi
gives
(b)
= K(x
term
of
+ x*
H----
+ x w + nc
)
which vanishes by
Hence, necessarily,
0(#)
which is Eq. (22), Art. 27.
5.
Therefore, integrating,
= Kx
follows from (a)
3>
of the
0,
(a).
= Kx,
Approximation Method for Observation Equations
Not
of the First Degree.

(Supplementary to Art. 39.)
This method requires that the values of the unknowns be
very approximately known beforehand, as by choosing

such of the observation equations as, when solved simul
taneously, will yield values for
all of
them.
Attention
is
then given to the unknown small corrections that must be

applied to these approximate values
a procedure some-
APPENDIX
what resembling Homer
179
method
approximation for
of
algebraic equations.
The approximate
ignated by ai, a 2
fy, and the corrections required by

u the true values of the unknowns are
,
q\, q
-, q
2,
however obtained, being des
values,
,
ft
Let the non-linear observation equations to be dealt with

be typified by
?2,
, qi)=s.
The
substitution of the values
/(oi
+ q\,
a2
+q
an observation equation
small
corrections q\,
member
of this
+g
/(cti
i,
a2
+ 5 ir~/( a
+ q\) = s,
+g
q\>
2,
OiH-g
a 2,
",
(e)
which the unknowns are the
Expanding the
by the general Taylor
i>
a. t
-~,
2,
in (d) gives
"
z,
in
(c)
+^
j)
theorem,
=/(ai, a 2
-/(oi,
ai)
02,
a
i
/fr\ a
being the remainder of the
higher powers
of
and which may

inaccuracy.
l>
series,
a 2?
**
>
therefore
neglected
&i)
+
4..B
T"
-t^>
which involves
the very small corrections

be
first
q\,
without
etc.,
serious
APPENDIX
180
f(ai, a2 ,
Denoting
by F,
a,)
by *,
aai
dF
byb,
00.1
we may
therefore replace
or
in
aq\
which
This
may
is
(e)
by
+ 6g +
2
+ rq
f
,
(/)
of the first degree,
and
F.
denotes s
an observation equation
be used as such, in combination with other observa
tions similarly obtained
from their respective
originals,
most probable values of the corrections.

This being done, the most probable values of the unknown
quantities themselves are found by adding the most
for finding the
probable corrections to the approximate values

h x2
f e~ dx.
a.
C.
Evaluation of the Integral
mentary
to Art. 54.)
This
may be
Equation
C"e-***dx.
Jo
transformed into
t/O
(70) is
(Supple-
APPENDIX
The new
hx
= z,
integral I
then hdx
181
independent of
is
h.
For,
dz,
Returning, however, to the original form of / in

multiply
e-dh
I e-
= f "Vu+^w
JQ
I and x being independent of
and
/*oo
e-*<to
rj[
Now
the
termined
hdh]dx.
we may
/&,
integrate both
of this equation with respect to h as
assigning the limits
tnUS
(g),
h
by e~ *dh:
it
members
let
and
/GO r /oo
integral within
a parameter,
to this integration also,
<x>
6 -a+^)w
the
brackets
dXf
(i)
is
readily de
2(1
hdh
^ C* e
o/
-V+*>
2(l+o; )/o
2(1
+z
ri
Substituting this in
(i)
gives
/*
l+o:
2-/o
7T
/00
But by Eq.
ber
is
(^)
the integral
equal to /
Hence I
is
equation (72).
flT
4
/I
VTT
which
^dh in the
first
*/o
and from
(a)
mem-
182
APPENDIX
D. Evaluation of the Probability Integral.

(Supple
The value of the integral
mentary to Art. 55.)
hX 1 by developing e~* into a

and integrating the terms separately. By Mac-
be found when
may
series
<
laurin s theorem,
whence
This
series
calculation of
is
hX less than or
converges rapidly for values of
equal to unity,
and may therefore be employed

in this case.
We may
divergent.
When hX
>
1,
write
Integrating successively by parts,
Zx
e~^
2x
&
c/
O r
A r3
^*Crt*t/
Of/
0^,5
O*o
Q
O
I
I
c/
C^
fa
^6
t
in the
however,
it
APPENDIX
183
+ 4X -A
e-l-^
2x J^
So:
3
3-5
16
3-5-7
,,v
..
64 x
Now
x2
I
Note
(See
(7.)
<r
d;r
e~*dx-
The value
of the integral in this last ex
pression can be found by applying the limits

oo to the successive terms of (k), giving
hX
and
00
2hX
3
15
J
which converges rapidly when
will
now
1.
^X>
Equation
(/)
F for
give values for the integral appearing in
this case.
Therefore, for
series
(ra)
values
of
hX<l,
use series
substituted in
the
(/).
probability
E.
will
(76)
integral
student verify these calculations
hX =
( j)
for,
for
hX>l,
use
then yield the
desired.
say,
Let
hX =
the
% and
2.
Outline of Another
Method
for Probable Errors
The
(Supplementary to Art. 62.)
method referred to at the end of Art. 62 is given here
of
Adjusted Values.
APPENDIX
184
(See Merriman, Method of Least Squares,
without proof.
Art. 74.)
Let the normal equations found from the observation

equations be
which the quantities K take the place of 2(aws), etc.,

equations (108). Let the literal form of these quanti
in
in
ties
K be preserved throughout the solution of the normal
equations, which will then yield
mi
o/fi
BoKo
-\-
-\
1-
^K,.
(n)
the quantities a,
the
|8,
literal quantities
It
X being numerical
coefficients of
K.
may then be shown
that the weight of mi
is
that of
ai
1
2
is
that of ms
1
is
etc.
That
is,
the weight of any
73
182
most probable value m p is the reciprocal of the coefficient

of the absolute term K p of the normal equation cor
responding to
mp
The weights
as
K p appears in the
of all the
solution (n) for
most probable values m\,
ra 2 ,
mp
The probable error of an obser

vation of unit weight is given by (107), or, as remarked
The
in Art. 62, may be already known from experience.
mi are thus calculated.
APPENDIX
probable error of each
m may
therefore
dividing this standard probable error

of the weight of
F.
185
now be found by
by the square root
above determined.
(92), as
Collection
of Important Definitions, Theorems,

for Convenient Reference.
Formulas
Rules and
DEFINITIONS
The
Error.
result of a
measurement minus the true
value of the quantity measured.

Residual.
The
result of a
most probable value

series of
(Art. 7.)
measurement minus the
of the quantity, as derived
measurements.
from a
(Art. 7.)
Most Probable Value.
calculated value of an
un
known quantity, based upon the results of measurements,

such that the residuals arising therefrom will be most
nearly in accord with the normal error distribution.
(Arts. 7, 29.)
Adjustment.
of
The
process of obtaining from the results
measurements the most probable values
quantities sought.
of the
unknown
(Chap. V.)
An
Observation Equation.
equation, in general only
approximately true, connecting one or more unknown

quantities, or functions of them, with the result of a
measurement.
(Art. 31.)
Normal Equation.
An
equation, in general one of a
set of simultaneous equations,

observation equations.
of the
whose solution gives the
unknowns involved
(Art. 33.)
in the
APPENDIX
186
An equation expressing
Equation of Condition.
a theo
which must be exactly satisfied by the cal

culated most probable values of the unknowns.
(Art. 40.)
retical condition
Empirical Formula.
formula expressing a relation
between variables, whose mathematical form is inferred

from the results of experience or experiment, and which
is
not deduced theoretically.
Numbers
Weights.
(Art. 42.)
assigned to observations, or to
the adjusted values of unknowns, representing the relative

degrees of confidence which the respective observations
or values are supposed to merit.
(Art. 47.)
The most probable value of a single

unknown quantity obtained by multiplying each obser
vation upon that quantity by its weight, adding the prod
ucts, and dividing by the sum of the weights.
(Art. 48.)
Weighted Mean.
Probable Error.
theoretical quantity
e,
so related
to the precision of a system of observations, that the
probability of the error of
any observation or adjusted
value being numerically
than
less
e is
equal to the proba
bility of its being numerically greater.
(Art. 58.)
RULES AND THEOREMS

Principle of Least Squares.
1.
(a)
The most prob
able value of a measured quantity that can be deduced
from a
and
skill, is
residuals
(b)
made with equal
series of direct observations,
is
that for which the

a
minimum.
sum
care
of the squares of the
(Art. 29.)
of
an unknown quantity
that can be deduced from a set of observations upon one
APPENDIX
of its functions is that for
the residuals
is
187
which the sum of the squares
minimum.
of
(Art. 31.)
The most probable values of unknown quantities

connected by observation equations are those for which
the sum of the squares of the residuals of those equations
is a minimum.
(Art. 33.)
(d) The most probable values of unknown quantities
(c)
connected by weighted observation equations are those for
which the sum
minimum.
2.
of the
weighted squares of the residuals
is
(Art. 52.)
Rules for
Adjusting
Observation
Equations of the
Write the expression for the residual

corresponding to each observation equation, multiply it
First Degree.
(a)
by the coefficient of the first unknown, in that

add the products, and equate their sum to
result
is
Do
likewise for each of the other
first
unknowns.
solve the normal equations thus formed for the
desired
(6)
The
the normal equation pertaining to the said
unknown.
Then
expression,
zero.
of the
unknowns.
(Art. 34.)
In the case of weighted observation equations, after
multiplying the residual by the coefficient of the unknown,
multiply again by the weight of the corresponding obser

vation
3.
then add and proceed as above stated.
Weight and Precision Index.
(Art. 48.)
The weights
of ob
servations are directly proportional to the squares of their

precision indices.
4.
(Art. 51.)
Weight and Probable Error.
The weights
of obser
vations are inversely proportional to the squares of their
probable errors.
(Art. 59.)
APPENDIX
188
FORMULAS
1.
The Error Equation.
(Art. 54.)
(74)
2.
(a)
Formulas for the Precision Index. (Arts. 57, 59.)

For observations of equal precision, standard for
mula,
(b)
For weighted observations, standard formula,

(93)
(c)
Peters
formula,
disregarding
signs
of
residuals,
observations not weighted,
n fr-
1}
(82)
Formulas for the Probable Errors of Observations in

Terms of Residuals. (Arts. 58, 59.)
3.
(a)
Probable error of single observation, no weights,
standard formula,
= 0.6745 X /-5
(6)
With weights
(88)
assigned, probable error of single ob
servation of unit weight, standard,
= 0.6745
A/^y-
APPENDIX
For an observation of unit weight, there being
(c)
known
Peters
(d)
and
= 0.6745
un
Simplified
(a)
when n
(a),
(c), disregarding signs of residuals,
^P
.
1 ^
= 0.8453-^=^=.
above
(107)
formulas corresponding to the above
= 0.8453
(e)
quantities, standard,
e
(6)
189
is
and
SQ x
(89)
(95)
= 0.8453
Peters
(6),
formulas
corresponding
to the
adapted to approximate calculation
large, disregarding signs of residuals,
= 0.85^.
n
(90)
(96)
4.
ties,
Formulas for Probable Errors of Functions of Quanti

in Terms of Probable Errors of Quantities Themselves.
(Art. 61.)
(a)
Function
of a single quantity q,
E = ê.
da
(97)
APPENDIX
190
(b)
Function
(c)
Function
Q=
E = VKi
of several quantities, q\, q%,
V + ^2
62
= KfJ
Function
...
qlt
+#
H^ + K q +
2
,V-
(103)
(104)
Formulas for Probable Errors of Adjusted Values.
5.
(Art. 62.)
(a)
For the arithmetical mean, standard,
OT
*p
= 0.6745 J
\rc(n
(6)
and
(105)
For the weighted mean, standard,
= 0.6745
(c)
\1)
Peters
(b)
formulas
corresponding to the above
(Ex. 10, Art. 64),

em
(106)
= 0.84S3
Sp
.
nVn-1
= 0.8453
Printed in the United States of America.
(a)
IE
following pages contain advertisements of a
few of the Macmillan books on kindred
subjects
The Mathematical Theory

and
Its
and
Statistical
Application
BY
to
of Probabilities
Frequency Curves
Methods
ARNE
FISHER,
F. S. S.
(London)
Cloth, 8vo, $2.00
For a number of years there has existed no adequate

the theory of probabilities in the English language,
who had wished
to study this fascinating
treatise
on
and the students
branch of applied mathe
matics were forced to turn to the treatises of various continental
The
writers.
this long-felt
present work by Mr. Fisher
want and place
in the
it is
hoped
will
I,
the theory of mathematical
ready, contains
up-to-
Volume
date and complete presentment of the probability theory.
now
remedy
hand of the student an
probabilities
proper and their application to that part of statistics which the au

thor has called homograde statistics.
especial stress
is
point
upon which
is
laid
the test of the stability of various statistical series
whether given as absolute or relative frequencies.
Numerous and
worked out examples based on American and European
carefully
statistical
data illustrate the theory.
critical attitude
taken to
many
feature of the
work
is
questions where often a mere
the
me
chanical use of mathematical formula without a thorough examina

tion of the philosophical aspects underlying such problems
may
to misleading or even paradoxical answers.

Publishers
64-66 Fifth Avenue
New York
lead
Fundamental Concepts
and Geometry
BY
of
Algebra
JOHN WESLEY YOUNG
Professor of Mathematics in Dartmouth College
The Growth of Algebraic Symbolism," by W. G.

MITCHELL, Professor of Mathematics in the University of Kansas.
With a note on
"
New
An
iqn.
247 pp.,
isrno, $f.6o
elementary account of the logical foundations of alge
bra and geometry,

cal
York,
elementary
in the
sense that the techni
mathematical equipment presupposed on the part of the
reader has been reduced to a minimum.
Except
in a very
few instances no knowledge of mathematics beyond the most

elementary portions of algebra and geometry has been as
sumed.
The author
gives a general exposition of the ab
stract formal point of
view developed during the
last
few
decades, rather than an exhaustive treatment of the details of

the investigations.
The
results of recent
work on the
logical
foundations are of vital interest alike to the teachers of math
ematics in our secondary schools and colleges and to philoso
phers and logicians. It is hoped that both these classes will

welcome a concise statement of some of the more fundamen
tal of
all
these results and an elementary exposition, free from
involved details, of the point of view which governs
present work on the foundations.

able also as a text in Teachers
,
The book should be

Courses
in
colleges
universities.

Publishers
64-66 Fifth Avenue
New York
all
avail
and
UNIVERSITY OF CALIFORNIA LIBRARY
BERKELEY
Return to desk from which borrowed.
This book is DUE on the last date stamped below.
ASTRDN
OCT 101
APfrdT
-19 S3
AUG 1 9 I960
JUt-31958
?
LD
21-100m-ll, 49(B7146sl6)476
YE
72965

Weld (Theory of Errors and Least Squares) 1916

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Weld (Theory of Errors and Least Squares) 1916

Загружено:

Авторское право:

Доступные форматы

GIFT

THEORY OF ERRORS AND

THE MACMILLAN COMPANY

THEORY OF ERRORS AND

PROFESSOR OF PHYSICS IN COE COLLEGE

THE MACMILLAN COMPANY

BY THE MACMILLAN COMPANY.

Published March, 1916.

Berwick & Smith Co.

wider applicability to general scientific work than the

which are capable of greater usefulness to the research

some reason, students are rarely given opportunity

acquire facility in these lines, the result being that too

and engineers go about their

work without such equipment.

chemist or the surveyor is daily called upon to do.

all, without the constant application

be gained from the application of the theory of pre

to this the fact that the

theory of error distribution has direct theoretical bear

and the reasons for students having opportunity to

intended as a presentation of the Theory

form as to be useful, not only as a textbook for

research worker can read through in an evening or so

and then put into immediate

be noticed that the illustrative examples and

employed by expert computers, often included in larger

thread of the subject, a few of the more complicated

Appendix and referred

to at the appropriate places.

It is not intended that they shall be omitted

when using the book as a text, though the cas

data for the illustrative examples, or

works, due credit has been given for the same.

CEDAR RAPIDS, IOWA,

impossibility of exact measurements

ON THE OCCURRENCE AND GENERAL

Errors and residuals

Mathematical expression of the law of error

Probability of either of two or more events

THE ERROR EQUATION AND THE PRINCIPLE

deduction of the error equation

principle of least squares in

ON THE ADJUSTMENT OF INDIRECT

More observations than

Reduction of observation equations of the

Observation equations not of first degree

Uses and limitations of empirical formulas

Relative reliability of observations. Weights

Wild or doubtful observations

General statement of the principle of least squares

PRECISION AND THE PROBABLE ERROR

Discontinuity of the error variable

Probability of an error lying between given limits.

Approximate formulas for the precision index

Relation between probable error and weight

Probable errors of adjusted values

A. Proof of the necessary functional relation assumed in

(Supplementary to Art. 39)

Evaluation of the integral

E. Outline of another method for probable errors of adjusted

F. Collection of important definitions, theorems, rules, and

THEORY OF ERRORS AND

THEORY OF ERRORS AND LEAST