You are on page 1of 6

# Method of Least square:

Assumptions:

1) The uncertainties are the same for all of the data points.

y=mx +c

## If point i, with co-ordinates xi and yi lies on the line, we have

y i=m xi +c

If the point does not lie on the line, the y-co-ordinate has error by distance e y i

So, y i=m xi +c +e y i

or, e y = y im xi c(i)
i

Now, according to the theory of least squares, the best straight line through a set of points is that
which minimizes e 2y i
i

## i.e. applying the least squares criterion e 2y i is minimum then

i

e2y i
i
=0(ii)
m

e2y
=0(iii) From equations (i, ii and iii), we
i
i

c
have normal equations

y i=m x i +nc(iv)
x

x
( i )+c x i(v)
2

( i y i )=m

Then the intercept c and the slope m can be found from Cramers rule

n ( x i y i ) x i y i
m=

c=
( x 2i ) y i x i (xi y i)

Where,
2
=n ( x 2i )( xi )

## The uncertainty is calculated as

2
m 2=n

2 2
( c ) =

( x 2i )

2 1 2
n2 i
with = ( y m x ic )

Class work

Example:

In an experiment to determine the acceleration of the free fall g the period T of a simple
pendulum was determined for various values of the l. The following results were obtained.

## Length (m) 0.25 0.35 0.45 0.55 0.65 0.75

Time (s) 0.98 1.21 1.34 1.50 1.60 1.75

Plot a graph of T2 against l, and use it to obtain a value for g, with the associated uncertainty.
Example:

In an experiment to determine the acceleration of the free fall g the period T of a simple
pendulum was determined for various values of the l. The following results were obtained.

## Length (m) 0.25 0.35 0.45 0.55 0.65 0.75

Time (s) 0.98 1.21 1.34 1.50 1.60 1.75

Plot a graph of T2 against l, and use it to obtain a value for g, with the associated uncertainty.

=== T =2
g
l

4 2
or, T 2 = ( )
g
l

## Comparing it with equation of straight line y = m.x + c; we have

4 2
y = T2 , m= ,l=x ,c=0
g

m=
( xi yi )
Since, c = 0,
( x 2i )
2

m 2=
( x 2i )

1
with 2= 2
n1 i
( y m x i)

mx = y-mx (y-mx)2
( xi y i )
l=x T T2 = y x.y x2
( x 2i )
x
0.25 0.98 0.9604 0.2401 0.0625 1.0088 -0.0484 0.002343
0.35 1.21 1.4641 0.51243 0.1225 1.41232 0.05178 0.002681
0.45 1.34 1.7956 0.80802 0.2025 1.81584 -0.02024 0.00041
0.55 1.50 2.2500 1.2375 0.3025 2.21936 0.03064 0.000939
0.65 1.60 2.5600 1.6640 0.4225 2.62288 -0.06288 0.003954
0.75 1.75 3.0625 2.29688 0.5625 3.0264 0.0361 0.001303
2
x i= y i= x i y i= 2
x i = ( y im x i) =

3 12.0926 6.75893 0.011629
1.675

m=
( xi yi )
Now, [since, ]
( x 2i )
6.75893
m= = 4.0352
1.675

## Thus, equation of best fitted line is

y = 4.0352 x x 0 1
y 0 4.0352
Also,

1
2 = 0.011629
61

= 0.0023258

= 0.05

m= 0.03726

= 0.04

4 2
Now, m=
g

or, g = 4 (3.14159)2/4.0352

= 9.7835

Hence,

## g = 9.78 0.04 ms-2

T2 Vs. l
3.5
3 f(x) = 4.04x
2.5
2
T2 (s2)

1.5 Linear ()
1
0.5
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
l (m)

Probable error:

(i)
R = R 1 + R2
Taking log on both sides
ln R = ln ( R1 + R2 )
Differentiating we get,
R R1 R2
= +
R R 1+ R 2 R 1+ R 2

## Then probable error in R is

2 1
R= R
[( ) ( ) ]
R1 2
R 1+ R 2
+
R2
R 1 + R2
2

(ii)
R 1 R2
R=
R 1 + R1
Taking log on both sides
ln R = ln (R1R2) - ln (R1+R2)
Differentiating we get,
R R1 R2 R1 R2
= +
R R1 R2 R 1+ R 2 R 1 + R 2

## Then probable error in R is

2 1
R= R
[( ) ( ) ( ) ( ) ]
R1 2 R 2 2
R1
+
R2
+
R1 2
R 1+ R 2
+
R2
R1 + R 2
2