Вы находитесь на странице: 1из 43

Classic Probability Theory

number of favorable outcomes


total number of possible outcomes
P(A) =
If A and B are independent events [occurrence of one has
no effect upon the occurrence of the other]:
P(A and B) = P(AB) = P(A)P(B)
probability of flipping two coins and having both come up
heads: 0.5*0.5 = 0.25, 25% chance
P(AB) = 0 when A and B are mutually exclusive
probability of a single coin flip resulting in both heads and
tails..
1. Probability of A and B occurrence [joint probability]:

2. Probability of A or B occurrence:
for mutually exclusive events, P(A+B) = P(A) + P(B)
probability of one club or one spade chosen in one draw from
a deck of cards = 13/52 + 13/52 = 0.5
For non-mutually-exclusive events, P(A+B) = P(A) + P(B)
P(AB)
avoids double-counting
probability of drawing either an ace or a spade on a given
draw from a deck of cards = 4/52 [an ace] + 13/52 [a spade]
1/52 [the ace of spades] = 0.3077
3. Conditional probability: probability of an event conditional
upon occurrence of another event.
) (
) (
) (
B P
AB P
B A P =
= ) ( B A P
probability that A will occur given that B has occurred
) (
) (
) (
A P
AB P
A B P =
) ( ) ( ) ( ) ( ) ( A P A B P B P B A P AB P = =
17,000 7,000 10,000
15,950 6,250 9,700 Acceptable
1,050 750 300 Defective
total Vendor 2 Vendor 1
Let:
A
1
= event part from vendor 1
A
2
= event part from vendor 2
B
1
= event acceptable part
B
2
= event defective part
event acceptable part from vendor 1
acceptable/defective parts from vendors 1 and 2
=
1 1
A B
59 . 0
17000
10000
) (
1
= = A P
and a few calculations.
41 . 0
17000
7000
) (
2
= = A P
97 . 0
10000
9700
) (
1 1
= = A B P 11 . 0
7000
750
) (
2 2
= = A B P 03 . 0
10000
300
) (
1 2
= = A B P
89 . 0
7000
6250
) (
2 1
= = A B P
Question: What is the probability of selecting a single part that is
made by vendor 1 and is also defective?
018 . 0 03 . 0 * 59 . 0 ) ( ) ( ) ( ) (
1 2 1 1 2 2 1
= = = = A B P A P A B P B A P
In a particular village, there are 60 women and 40 men. Twenty of
those women are 70 years of age or older; 5 of the men are 70
years of age or older.
Example 2 [Cliffs Quick Review of Statistics]
What is the probability that a person selected at random in that
town will be a woman? P(A
1
) = 60/100 = 0.60
What is the probability that a person 70+ years of age selected
at random will be a woman? = 20/25= 0.80
A
1
= event woman
A
2
= event man
B
1
= event woman, 70 or older
B
2
= event man, 70 or older
=
2 1
A B
event 70 or older and woman
******************************************************
) B A P(
1 1
What is the probability that a person selected from the
village population will be a woman, 70 years of age or
older?
Finally..
2 . 0 ) 60 / 20 ( * 60 . 0 ) ( * ) ( ) ( ) (
1 1 1 1 1 1 1
= = = = A B P A P A B P B A P
Probability Distributions the Binomial
experiment consists of a number of identical events (n)
each event has only one of two mutually exclusive outcomes
[success or failure]
probability of success outcome equal to some percentage,
identified as a proportion, t,
t remains constant throughout all events, defined by ratio of
number of successes to number of trials
events are independent
( )
( )
x n
x
x n x
n
x P

= t t 1
)! !
!
) (
n! = n(n-1)(n-2)(3)(2)(1)
Example: A coin is flipped 10 times. What is the
probability of getting exactly 5 heads?
n = 10
x = 5
t = 0.5
( )( ) 246 . 0 5 . 0 1 5 . 0
) ! 5 ( ! 5
! 10
) (
5
5
= = x P
Callister, 4
th
Edition, Problem 6.50
Below are tabulated a number of Rockwell B hardness values,
which were measured on a single steel specimen. Compute the
average [mean] and standard deviation for the measured
hardness values.
86.3 85.5 87.2
84.4 83.5 86.2
86.9 87.8 82.8
85.2 84.7 88.3
86.4 80.7 83.3
3 . 85
1
= =

=
n
x f
x
n
i
i i
08 . 2
1
) (
2 / 1
1
2
=
(
(
(
(

=

=
n
x x
s
n
i
i
Callister, 4
th
Edition, Problem 6.50
. and a couple of other things.
variance = s
2
= 4.33 [a measure of dispersion of sample]
coefficient of variation, v, =
arithmetic mean for sample = , standard deviation = s
arithmetic mean for population = , standard deviation = o
x s /
x
Callister, 4
th
Edition, Problem 6.50
81 82 83 84 85 86 87 88 89
Rockwell B Hardness
0
5
10
15
20
25
30
F
r
e
q
u
e
n
c
y
,

%
80 81 82 83 84 85 86 87 88 89
Rockwell B Hardness
0
5
10
15
20
25
30
F
r
e
q
u
e
n
c
y
,

%
Probability that a
single hardness
measurement
will result in a
value greater
than R
B
87
Callister, 4
th
Edition, Problem 6.50
The Normal Distribution
(
(

|
.
|

\
|

=
2
2
1
exp
2
1
) (
o

t o
x
x f
80 81 82 83 84 85 86 87 88 89
Rockwell B Hardness
0.00
0.05
0.10
0.15
0.20
0.25
F
r
e
q
u
e
n
c
y
,

%
f(x) = height
of frequency
curve at
assigned value
for x [RB
value]
o = standard
deviation of
population
= mean of
population
-4 -3 -2 -1 0 1 2 3 4
z=(x-u)/s
0
0.1
0.2
0.3
0.4
0.5
r
e
l
a
l
t
i
v
e

f
r
e
q
u
e
n
c
y

[
z
]
The Standardized Normal Frequency Distribution
The Standardized Normal Frequency Distribution
z = [x ]/o
From SND [standard normal
distribution tables], can read off the
area under the z-distribution curve at
a particular value for z
For z = -1, 15.87% of the samples
represented by the z-distribution have
values of z which are less than 1.
The Standardized Normal Frequency Distribution
2 1 0 1 2
standard normal variable, z
0.00
0.05
0.10
0.15
0.20
0.25
p
r
o
b
a
b
i
l
i
t
y

o
f

z
,

p
(
z
)

For the
Callister case,
where the
mean and
standard
deviation are
85.3 and 2.08,
respectively, x
= 83.2
15.87% of
measured R
B

[~2/15] are
less than 83.2
Weibull Distributions [positive values e.g. fatigue lives]
(
(

|
.
|

\
|

|
.
|

\
|
=
m m
x x m
x f
u u u
exp ) (
1
Steering Arm Forging Defects
m = shape parameter,
Weibull Modulus
[Weibull Slope]
theta, u, scale
parameter or
characteristic value
For oxide scale 63.2%
of forgings have
thickness values within
this range [0.6 mm]
Weibull Distributions [positive values e.g. fatigue lives]
o
max
of 903.2 MPa
during turning
threshold of 8.8
MPa*m
1/2
for
fatigue crack
propagation
any flaw larger
than 0.07 mm will
propagate
Weibull Distributions [positive values e.g. fatigue lives]
@ 12 hard turns
per day, 365
days/year for 3
years, accumulate
13,000 cycles
Any control arm
with a flaw
greater than ~ 0.9
mm would be
expected to fail in
3 years
How does the Distribution affect Samples drawn from it?
Distribution of Sample Means [standard error of the mean]
n
x
o
o =
n
x x
z
x
/ o

o

=

=
t Distribution [when o for the population unknown]
n s
x
t
/

= for v = n 1 degrees of freedom
F Distribution
2
2
2
1
,
2 1
s
s
F =
v v
Determines whether samples come from
populations having equal variances
What does a samples or series of samples tell us about the
population that the sample came from?
Establish null and alternative hypotheses
H
o
: =
o
H
1
: <
o
, >
o
Procedure is to propose an appropriate hypothesis and its
alternative before the experiment is conducted.
In allowing for some variation between the two populations, two
types of errors arise:
Type I : probability of Type I = o, where null rejected although true
Type II: probability of Type II = |, where null accepted although
false
Select test statistic
Select test criteria [o and |] and sample size [n]
Obtain sample and compute sample characteristics
Determine critical region for rejecting H
o
Make decision
Continuing the decision making process.
property value
0
0.01
0.02
0.03
0.04
0.05
0.06
f
r
e
q
u
e
n
c
y
decision
point
o
o
|
leaded red/semi-red brasses
Introduction

Federal regulations now limit the amount of lead permitted in public
drinking water supplies. These regulations, in turn, established an
immediate need for a reduced lead or lead-free plumbing alloy. The
Environmental Protection Agency's Lead and Copper Rule applies to
public water utilities and establishes a framework to monitor and
control water chemistry in order to reduce the water's corrosive
power. Utilities are required to chemically treat water that is
aggressive enough to leach excessive copper or lead from the water
supply system. The EPA rule reduces lead "action levels" (at which
appropriate water treatment steps must be applied) from 50 to 15 parts
per billion (ppb), as measured at the household tap, in at least 90% of
households.
leaded red/semi-red brasses
The National Sanitation Foundation's NSF Standard 61 uses the U.S.
EPA's Lead and Copper Rule action levels, and its Maximum
Contaminant Levels for other substances, to establish standards for
contaminant pickup from equipment and hardware of a water supply
system. Faucets to be certified under NSF Standard 61 must pass a test
in which representative products are filled with an aggressive pH 8 test
water. Samples of the water are withdrawn and analyzed for lead at
specified intervals over a 19-day period. Statistical tests are applied to
the data to ensure with 90% confidence that 75% of the products
examined leach no more than 11 g of lead into the water. (The 11 g
limit under NSF Standard 61 is lower than the 15 g of the Lead and
Copper Rule because EPA assumes that as much as 4 pg of lead could
be picked up from other sources.) The recent California Proposition 65
consent judgment goes even further: it calls for a 5 g limit using the
same NSF Standard 61 test protocol.
Los Angeles Department of Water and Power [LADP]
Storage Facility
typical curb
stop valve
used by EQI
in their
analysis.
Note that
band saw
used to
produce
shavings
which were
analyzed by
EDX
metal analysis site
ball valve
Typical curb
stop ball
valve
assembly
interior.
Note valve
has at least
two curb
components:
ball valve
and valve
body
valve body
Over the entire service life of a leaded brass plumbing
component, it would be predicted that higher lead content
brass alloys would dissolve out a greater total mass of lead.
Specifically, it would be expected that initially a 5-percent
leaded brass alloy and a 7-percent lead alloy might leach
similar concentrations of lead when there is an excess of lead
present on the surface of each alloy.
However, as this lead layer is preferentially dissolved, the
amount of surface lead exposed would become a limiting factor
for the 5-percent alloy before it becomes a factor for the 7-
percent alloy.
Galvanic Coupling in Curb Stop Ball Valves
C836 85 metal
C844 81 Metal
C360 60-40 Brass
dispersed
lead phase
copper-tin-zinc
solid solution
matrix
0 10 20 30 40 50 60 70 80 90 100
0
200
400
600
800
1000
1200
1400
Composition, weight percent lead[w/oPb]
T
e
m
p
e
r a
t u
r e
, C
Cu
liquid
L
1
L
2
o+L
L(i)+L(ii)
955C 36 87
326C
o+|
o+L
99.4
991C, 63w/oPb
Pb concentration in alloy
Pb concentration in
last liquid to freeze
Galvanic Coupling in Curb Stop Ball Valves
Table I - Bulk Alloy and Phase Compositions for Leaded Plumbing Alloys
- - 0 100 0 0 lead [Pb]
42.0 - 58.0 36.6 - 63.4 35.5 3 0 61.5 C360
11.8 3.8 84.4 9.7 3.2 87.1 9 7 3 81 C844
6.5 6.3 87.2 5.3 5.3 89.4 5 5 5 85 C836
Zn Sn Cu Zn Sn Cu Zn Pb Sn Cu
matrix composition
volume percent
matrix composition
weight percent
alloy composition
weight percent
alloy
designation
(

=

a
c
c a
A
A
j j Please remember:
0.401 O
2
+ 2H
2
O + 4e
-
= 4OH
-
-0.126 Pb
+2
+ 2e
-
= Pb
-0.136 Sn
+2
+ 2e
-
= Sn
-0.763 Zn
+2
+ 2e
-
= Zn
0.334 Cu
+2
+ 2e
-
= Cu
standard potential, E
o
, volts SHE process
30.1(33.7) 13.7(8.2) 11.5(12.6) 7.5(4.7) 27 7 MWD
17.7(6.5) 10.0(3.9) 7.2(2.8) 5.4(2.2) 13 5 MWD
20.0(2.6) 8.3(1.9) 7.0(0.4) 4.9(0.4) 9 7 LAA
16.2(5.6) 8.1(2.6) 6.4(1.6) 4.6(1.0) 10 5 LAA
mean(sd) mean(sd) mean(sd) mean(sd)
16-hr 1-hr 30-min 10-min # of
meters
Pb
%
water
type
Pb Leaching from Leaded Plumbing Valves
2 . 16
1
= x
s
1
= 5.6
n
1
= 10
v
1
= 9
0 . 20
2
= x
s
2
= 2.6
n
2
= 9
v
2
= 8
Consider the quantity of lead
leached into water during a 16 hour
interval [~ 2*overnight]
0 5 10 15 20 25 30 35
Pb Concentrations in Leachate, microgram/L
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
f
r
e
q
u
e
n
c
y
7% Pb
5% Pb
2 . 16
1
= x
s
1
= 5.6
normal distribution plots of 16 hour valve leaching experiments
0 . 20
2
= x
s
2
= 2.6
Los Angeles
Aquaduct
[LAA] water
used for
leaching
experiments
Pb Leaching from Leaded Plumbing Valves
F = s
1
2
/s
2
2
=
5.6
2
/2.6
2
= 4.639
Critical region
determined from
F distribution
chart [Table B-
3]:
5.91
o
=0.01
3.39
o
=0.05
2.56
o
=0.10
v2
= 8
v1
= 9
Pb Leaching from Leaded Plumbing Valves
critical region of F for o = 0.05 lies outside of t
o/2
<t<t
o/2
, t +/- 3.39
since F = 4.639 > 3.39, we can assume that the samples
came from populations of unequal variances
= and o
1
o
2
entry 7 from the

statistical test of hypotheses table [Table
10-3] is used
hypothesis, H
o
:
1

2
= o = 0
conditions: o unknown, o
1
not equal to o
2
distribution of test statistic: t distribution
alternate hypothesis, H
1
:
1

2
not equal to o [0]
Pb Leaching from Leaded Plumbing Valves
927 . 1
9
6 . 2
10
6 . 5
0 ) 20 2 . 16 ( ) (
2 / 1
2 2
2 / 1
2
2
2
1
2
1
2 1
=
|
|
.
|

\
|
+

=
|
|
.
|

\
|
+

=
n
s
n
s
x x
t
o
test statistic
distribution of test statistic
( )
( )
13 ~ 984 . 12
8
807 . 0 1
9
807 . 0
1
1
1
2
2
2
2
1
2
=
|
|
.
|

\
|

+
=

+
|
|
.
|

\
|
=
v v
v
k k
807 . 0
9 / 6 . 2 10 6 . 5
10 / 6 . 5
/ /
/
2 2
2
2
2
2 1
2
1
1
2
1
=
+
=
+
=
n s n s
n s
k
Pb Leaching from Leaded Plumbing Valves
3.646 2.898 2.567 2.110 1.740 1.333 17
3.852 3.012 2.650 2.160 1.771 1.350
13
4.297 3.250 2.821 2.262 1.833 1.383 9
0.002 0.010 0.020 0.050 0.100 0.200
v
two-tail
o
0.001 0.005 0.010 0.025 0.050 0.100
one-tail
o
From the t distribution table [Table B-2]
The critical value for t in this test would be the one-tail o
value since we are looking to ascertain whether there is a
statistical difference in Pb leached from the two collections of
meters specifically, that
2
>
1
, with a 95% confidence level
Pb Leaching from Leaded Plumbing Valves
Since the computed t-value for a o of 0 [-1.927] is less than
the critical value [- 1.771], the null hypothesis can be
rejected and one can be 95% confident that the alloy
containing 7% Pb will leach more lead than will the 5% Pb
alloy.
But can the same be said for the other residence times?
and, will 7% Pb alloy leach 100*[7-5]/5 = 40% more Pb than
the 5% Pb alloy i.e. is o large enough?
Assuming the sample valves had been conditioned during
service, can differences be ascribed to surface enrichment of
Pb?
Perhaps EQI had this in mind [????] when they concluded,
However, because of the small sample size and large
variability between meters, the differences between the 5-
percent and 7-percent results for each individual dwell time
are not statistically significant at p-values of 0.05.
Pb Leaching from Leaded Plumbing Valves
0.12 -5.6 difference
15.7 [12.8] 27 7% MWD
10.1 [3.8] 13 5% MWD
0.01 -1.2 difference
10.0 [1.0] 9 7% LAA
8.8 [2.6] 10 5% LAA
two-tailed P-value
for difference*
composited mean
1
[SD]
# of
meters
lead alloy
water
type
However, the chances of all the dwell time and water combinations
exhibiting higher lead concentrations for the 7-percent lead alloy
meters, if in fact no real differences exist, is much more unlikely. To
test this hypothesis more effectively, we composited the results of the
four standing times for all the meters of a given water and alloy.
F = o
1
2
/o
2
2
= 2.6
2
/1.0
2
= 6.76; for v
1
= 10-1 = 9, v
2
= 9-1 = 8,
critical region is F>3.39 for o=0.05. Therefore, o
1
not equal to
o
2
and entry 7 equation used.
H
o
:
1
-
2
= o = 0; H
1
:
1
-
2
= o, o not equal to 0
*****************************************************
859 . 0
9 / 0 . 1 10 6 . 2
10 / 6 . 2
/ /
/
2 2
2
2
2
2 1
2
1
1
2
1
=
+
=
+
=
n s n s
n s
k
( )
( )
12 ~ 84 . 11
8
859 . 0 1
9
859 . 0
1
1
1
2
2
2
2
1
2
=
|
|
.
|

\
|

+
=

+
|
|
.
|

\
|
=
v v
v
k k
distribution of test statistic
for one-tail o, critical region outside limits of t +/- 1.781
352 . 1
9
0 . 1
10
6 . 2
0 ) 10 8 . 8 ( ) (
2 / 1
2 2
2 / 1
2
2
2
1
2
1
2 1
=
|
|
.
|

\
|
+

=
|
|
.
|

\
|
+

=
n
s
n
s
x x
t
o
test statistic
and your conclusions are..?

Вам также может понравиться