Estimation of Parameters

Introduction
• Estimation is the process by which we guess or
estimate the population mean or variance from
the sample statistics.
• Estimation is made under two options:
1. Point Estimate. 2. Interval Estimate.
• A point estimate is a single valued estimate of
population parameter. Thus a sample mean is
used as an estimate of a population mean.
• Sample standard deviation is used as an estimate
of the population standard deviation.
 Introduction
• Where as an Interval Estimate is the range of
values within which the value of the parameter is
expected to lie, with a known risk of error.
• Hence, the population mean (  ) and Standard
deviation ( ) is known as parameters, while
sample mean ( x ) and sample S.D (s) is known as
Sample Statistics.
Population Sample
Parameters Statistics

mean(  ) mean( x )
S .D( ) S .D ( s )
 Introduction
• For example a 95% confidence interval would
carry with it a 5% risk of error that is, 5% of
interval so formed would not include the
population mean. Similarly other confidence
intervals would be defined.
Desired Confidence
(1-  )100% z /2 Formula

90% 1.65 
x  1.65
n
95% 1.96 
x  1.96
n
99% 2.58 
x  2.58
n
 Estimating the mean of a Population
• The method of estimating the mean of
population depends on whether the standard
deviation of population  ( ) is known or
 30 (
unknown and the selection of sample isnlarge
) or small (n<30), where as we expect that the
samples are drawn from normal population and
the sampling distribution of x to be normally
distributed with mean,

 x   and  x 
n
 Confidence Interval for the population mean

Case-I: If  is known and n  30 or n  30 .

     
x  z /2      x  z /2  
 n  n
Case-II: If  is unknown and n  30 , then replace
 by s.
 s   s 
x  z /2      x  z /2  
 n  n
 Confidence Interval for the population mean

Case-III: If  is unknown and n  30 ; then the case

is referred as a special case.

 s   s 
x  t      x  t  
2
, n 1
 n 2
, n 1
 n
 The T-distribution
• Most of the time we are not fortunate enough to
know the variance of the population from which we
select our random samples. For example of sample
size  30 estimate of
, angood is provided by
2
2
calculating . s
• When the population’s standard deviation  is
unknown and the sample 2
is small (less than 30),
then the values of s fluctuate considerably from
sample
x 
to sample and distribution of values
S/ n
is no longer
a standard normal distribution.
 The T-distribution
• We are now dealing with the distribution of a
statistic which we call “t” distribution whose
value is given by:
x 
t
S/ n
• In deriving the equation of this distribution, we
assume that the samples were selected from a
normal population. Note that the only
computational difference between z and t is the
substitution of S for  .
 The T-distribution
• Remember that 
is constant where as S is a variable
whose values depends upon a particular sample.
The distribution of t is similar to the distribution of
Z. In fact they both are symmetrical about a mean
of zero. Both distributions are bell shaped, but the
variation of t values is more than that of z values.
Knowing the fact that the t value depends
x and S2 on the
two quantities, . Wherex the z values
depend only in the changes S x from sample to
sample. If n is large, then will be small and hence
t-distribution curve will take a shape of normal
curve.
 The T-distribution
• For the Z statistic the normal distribution
provides the approximate model. For the t
statistic there is not one curve (distribution) but
infinitely many t curves; the one that is used
depends on the sample size. If the sample size is
n, then we identify the curve in question by
saying that it is the t-curve with n-1 degrees of
freedom. Therefore, we define degrees of
freedom simply as a number that identifies the
appropriate t-curve or distribution.
 Example#01
• A random sample of 49 observations is drawn
from a normal population with unknown mean
and variance. The sample mean is 131.8 and the
standard deviation 6.25. Setup a 90% confidence
interval for the population mean.
Solution:

In this problem is unknown and n>30,

 replace
therefore we by the standard deviation
S.
 Example#01
The 90% confidence interval is:
 s   s 
x  z /2      x  z /2  
 n  n
 6.25   6.25 
131.8  1.65      131.8  1.65  
 49   49 
130.33    133.27
Thus, we are 90% confident that the population
mean will lie in the interval (130.33, 133.27).
 Example#02
• A random sample of 20 observations is drawn
from a normal population with unknown mean
and variance. The sample mean is 131.8 and the
standard deviation 6.25. Setup a 90% confidence
interval for the population mean.
Solution:

In this problem is unknown and n<30,

therefore a special case-III is followed as:
 Example#02
The 90% confidence interval is:
 s   s 
x  t      x  t   
2
, n 1
 n 2
, n 1
 n
 6.25   6.25 
131.8  1.729      131.8  1.729  
 49   49 
130.256    133.344
Thus, we are 90% confident that the population
mean will lie in the interval (130.256, 133.344).
 Confidence Interval for estimating the
difference of two means ( 1  )2
Case-I:
If  1 and  2 are known and n1 & n2 are  30 or  30, then :
2 2

 
2 2
 
2 2
 x1  x2   z 1
  1  2   x1  x2   z
2

1 2

2
n1 n2 2
n1 n2
Case-II: If  12 and  22 areunknown and n1 & n2 are  30, then
replace  12 and  22 by S12 & S 22 ,

2 2 2 2
S S S S
 x1  x2   z 1
  1  2   x1  x2   z
2

1 2

2
n1 n2 2
n1 n2
 Confidence Interval for estimating the
difference of two means ( 1  )2
Case-III (a):
If  12 and  22 areunknownbut assumed  12   22  and n1 & n2 are  30, then :
1 1 1 1
 x1  x2   t ,v S p   1  2   x1  x2   t S p 
2
n1 n2 2
,v n1 n2
Sp
Where, is the pooled estimate of the population standard
deviation, given by:
( n  1) S 2
 ( n  1) S 2
S p2  1 1 2 2
n1  n2  2
t /2,v
And is the t-value with V degrees of freedom is given
by: v  d  f  n1  n2  2
 Confidence Interval for estimating the
difference of two means ( 1  )2
Case-III(b):
If  12 and  22 areunknownbut assumed  12   22  and n1 & n2 are  30, then :
2 2 2 2
S S S S
 x1  x2   t ,v 1
  1  2   x1  x2   t
2
 1 2

2
n1 n2 2
,v n
1 n2
Where, t /2,v is the t-value with
S / n1  S / n2 
2
2 2
1 2
v
S / n1  /  n1  1   S / n2  /  n2  1
2 2
2 2
1 2

Degrees of freedom, leaving an area  / 2 to the

right.
 Example
A sample of 100 cables was taken from a company
A and average breaking strength was found to be
1980 pounds with a standard deviation of 180
pounds, while another sample of 125 cables was
taken from another company B and the average
breaking strength was found to be 1950 pounds with
a standard deviation of 150 pounds. Find a 99%
confidence interval for the difference ,where
1  2 1
is the mean breaking strength of all cables of
2
company A and is the mean breaking strength of
all cables of company B.
 Example
The point estimate of
1   2 is x1  x2  1980  1950  30 pounds
n1 and n2
Since are both large, we can substitute
S1  180 and S 2  150 for  1 and  2 respectively
z /2  2.58, n1  100 and n2  125
Now,
  0.01
Using that is 99% confidence interval is:
 Example

S12 S22 S12 S22

 x1  x2   z   1  2   x1  x2   z 
2
n1 n2 2
n1 n2
 27.92  1  2  87.92

This shows that difference between breaking

strength of all cables produced by Company A
and Company B is (-27.92 to 87.92) with 99%
confidence.
 Estimating the variance
• If a sample of size n is drawn from a normal
population with variance  2 and the sample
variance S 2 is computed. We obtain a value of
the statistic
S2 . This computed sample variance S 2(
) will be used as a point estimate of 2
.
Hence the statistic S 2 is called an estimator of  2 .
An interval estimate of  2 can be estimated by
using the statistic:
(n  1)S
2
 
2

2
Called chi-square with (v  n  1) degrees of freedom.
 Confidence Interval for 
2

• If S 2 is the variance of the random sample of size

n from a normal population, a (1   )100 %
confidence interval for  is given by:
2

 n  1 S 2

 

2 n  1 S 2

 /2
2
1 /2
2

Where  /2 and 
2

2 2
are values with degrees
1 /2
of freedom leaving an areas of / 2 and 1   / 2
respectively to the right.
 Example
A manufacturer of a car batteries claims that the
batteries will last, on the average, 3 years with a
variance of 1 year. If 5 of these batteries have
life times of 1.9, 2.4, 3.0, 3.5 and 4.2 years,
construct a 95% confidence interval for  and
decide if the manufacturer’s claim that   1is 2

valid. Assume the population of battery lives to

be approximately normally distributed.
Solution: Variance of the sample S 2 is defined as ,
 x  x 
2
3.26
S 2
   0.81
n 1 5 1
 Example
X x  x  x  x 
2

1.9 -1.1 1.21

2.4 -0.6 0.36
3.0 0 0
3.5 0.5 0.25
4.2 1.2 1.44
Total 3.26
 Example
Using   0.05, v  4
we find  0.025  11.143 and  0.975  0.484
2 2

Using the formula:

 n  1  2  
S 2
n  1  S 2

 / 2
2
1 / 2
2

4  0.81 4  0.81
 2

11.143 0.484
0.29    6.69
2

Hence, the manufacturer claim is valid.

Confidence Interval for /  2
1
2
2
2 2
S
If 1 and S 2 are the variances of independent
samplesnof 1 and
size n2 respectively from the normal
populations, then(1   )100% confidence
1 /  2
2 2
interval for is:
S 2
1  S 2 2
1
  f /2  v2 , v1 
1 1
S f /2  v1 , v2  
2
2
S 2
2 2
2

f /2  v1 , v2  v1  n1  1 and v2  n2  1
Where is an f-value with  /2
/2  v2 , v1  leaving an area of
degrees offfreedom v2to nthe
2 1
 n1  1
rightv1and is a similar f-value with
and degrees of freedom.
 Example
To complete a certain job, a sample of 16
workers in plant-I took an average of 80 minutes
with a standard deviation of 4 minutes. A sample
of 11 workers in plant-II took an average of 75
minutes with a standard deviation of 3.5

minutes. Find a 98% confidence interval for 1 2 2
/  2

where  1 is the variance of all workers of plant-

2

I and  2 is the variance of time taken by plant-II.

2

Solution: We have n1  16, n2  11

S1  4 and S 2  3.5
For a 98% confidence interval
 Example
  0.02, using the table, wefind that
f 0.01 (15,10)  4.56 and f 0.01 (10,15)  3.80
Substituting in the formula:
S12 1  12 S12
 2  2 f /2  v2 , v1 
S 2 f /2  v1 , v2   2 S 2
2

we obtain 98% confidence interval

16 / 12.225 1/ 4.56    /   16 / 12.25 3.80 
2
1
2
2

1.306  0.219  1 2  1.306 3.80 

  2
/  2

0.286   12 /  22  4.962