Sampling Distribution and

Estimation:

@ Dinesh Shrestha 1
Parameter and statistic:
 The statistical measures or constants that are computed from the population
are called population parameters.
 Population mean ( X ) or µ, population variance (σ2), skewness (β1), kurtosis
(β2), population correlation (ρ) etc are some example of population
parameters.
 On the other hand, the statistical measures or tools that are computed from
the sample are called statistics.
 x
For example; sample mean ( ), sample variance ( s2), sample correlation(r)
etc.
 Parameter is the function of the population while the statistic is the
function of sample. Thus, the value of parameter is always constant while the
value of statistic is variable, since it depends upon the size of the sample.
 Average of statistic = Parameter
 Expectation of ( x ) = Parameter
 Or, E ( x ) = X or µ
 Similarly, E (s2) = σ2

@ Dinesh Shrestha 2
Some parameters and statistic
Statistic Parameters
 1. Sample mean: 1. Population mean.
x=
 x
X ( ) =
X
n N
1. Sample variance (biased). 1. Population variance
 ( x − x )2
(biased).
s2 =  ( X −  )2
n  =
2
1. Sample variance N
(unbiased). 1. Population
 (x − x )
2
variance(unbiased).
s2 =
n −1  ( X −  )2

1. Sample correlation. S2 =
N −1
Cov.( x, y ) 1. Population correlation.
r=
var( x) var( y ) Cov.( X , Y )
=
var( X ) var(Y )

@ Dinesh Shrestha 3
Sampling Distribution:
 Sampling distribution is defined as the probability distribution of
given statistic (measure of sample) based on a random sample.
 It is the probability distribution formed by the values of means or
variance or proportion of random samples selected from a given
population.
 The sampling distribution depends on the distribution of the
population, the size of the sample, and the methods of the
sampling.
 Without knowing the details about the sampling distribution of
corresponding statistic, it will be very difficult to draw a conclusion
about the unknown population parameters using sample data.
 Sampling distribution is very important in statistics because they
provide a major simplification on the way to statistical inference.
 It is more useful in estimation and inference analysis.

@ Dinesh Shrestha 4
Standard Error (SE):
 The standard deviation of the sampling distribution of a
statistic is known as standard error (S.E.)
 Standard error measures the reliability in sampling due
to chance. The reciprocal of the standard error is a
measure of precision or reliability, thus it is also called an
index of precision of the estimate of the parameter.
 Grater the standard error, lesser will be the precision of
the sample and lesser the standard error, greater is the
precision of the sample.
 It is used as a tool in estimation and hypothesis testing.
Thus, S.E. plays a vital role in the sampling theory.

@ Dinesh Shrestha 5
Sampling distribution of mean:
 The distribution formed from the values of means of
all the possible random samples selected from a
given population is called sampling distribution of
the sample mean.
 If the given finite population is normal distribution
then the sampling distribution of mean so obtained
is called sampling distribution of mean of random
sample drawn from the normal population.
 The sampling distribution of mean is used to
estimate unknown population mean in inferential
statistics.
@ Dinesh Shrestha 6
Properties of the sampling distribution
of mean:
 The main properties are,
1. Sample mean (𝑥)ҧ is an unbiased estimate of the
population mean (μ).
2. The variance of sample mean (𝑥)ҧ depends on the
sample size (n).

@ Dinesh Shrestha 7
Variance and standard error of the
sample mean (for single mean):
 Let 𝑥1, 𝑥2, … , 𝑥𝑛 be the random samples of size ‘n’
drawn from the population with mean (μ) and
variance (𝜎 2 ).
 Variance of the sample mean in case of without
replacement is given by
 V(𝑥)ҧ =   N − n  if σ is given.
2

n  N −1 
and standard error of the sample mean
  N −n
 S.E. (𝑥)ҧ = 
N −1


n 

@ Dinesh Shrestha 8
 If N is not given/known or infinite, then
2 𝜎
 V(𝑥)ҧ = S.E. (𝑥)ҧ =
n 𝑛
( x − x )
2

 Variance of sample mean (𝜎𝑥2ҧ ) = if samples

k
 Where, are given
 n = sample size N = population size
2 1
 Population variance (𝜎 ) = N  ( X −  )
2
=X =  X
N
 𝑥ҧ = sample mean
 k = sample number = N Cn x=
 x = mean of the
k sample means

@ Dinesh Shrestha 9
 When population standard deviation (σ) is not
known, then the standard error of mean is estimated
by using Root mean square value or unbiased sample
standard deviation (S).

 N −n
 S.E. (𝑥)ҧ = S   if N is known/finite.
n  N 

S
 S.E. (𝑥)ҧ = if N is unknown/infinite.
n
 Where,
1
S=  (x − x)
2
 if data are given
n −1

@ Dinesh Shrestha 10
 If biased sample standard deviation is given, then
the standard error of the sample mean is,

s  N −n
 S.E. (𝑥)ҧ =   if N is known/finite.
n −1  N 

s
 S.E. (𝑥)ҧ = if N is unknown/infinite.
n −1
 Where,
1
s=  (x − x)
2

@ Dinesh Shrestha 11
 Variance of the sample mean in case of with
replacement is given by
 V(𝑥)ҧ =  2
S.E. (𝑥)ҧ =
𝜎
n 𝑛
 Where,
 n = sample size.
=X =
1 X
 Population variance (𝜎 2 ) = ( X −  )
2

N
N
(𝜎𝑥2ҧ ) ( x − x )
2
 Sample variance of sample mean =
k
x=
 x
= mean of the
 𝑥ҧ = sample mean k sample means.
 k = sample number = N n

@ Dinesh Shrestha 12
Standard error for double means:
 If 𝑥ҧ1 and 𝑥ҧ2 be the sample means of sizes 𝑛1 and 𝑛2 then the
standard error of difference between two sample means is
given by
 S.E.(𝑥ҧ1 - 𝑥ҧ2 ) =  12  22 if 𝜎1 and 𝜎2 are given/known.
+
n1 n2

1 1
 S.E.(𝑥ҧ1 - 𝑥ҧ2 ) = 𝑆 2 𝑛1
+
𝑛2
if 𝜎1 and 𝜎2 are not given/unknown.

 Where,
  ( x1 − x1 )2 +  ( x2 − x2 )2 
 S2 =   if data are given
n1 + n2 − 2
n1s12 + n2 s22
 S =
2
if 𝑠1 and 𝑠2 are given.
n1 + n2 − 2
@ Dinesh Shrestha 13
Sampling distribution of proportion:
 The distribution formed from the values of
proportions of all the possible random samples
selected from a given population is called sampling
distribution of the sample proportion.
 The sampling distribution of proportion is used to
estimate unknown population proportion in
inferential statistics. The proportion represents the
ratio of the attributes to the total.

@ Dinesh Shrestha 14
Properties of the sampling distribution
of proportion:
 The main properties are,
1. Sample proportion (p) is an unbiased estimate of
the population proportion (P).
2. The variance of sample proportion (p) depends on
the sample size (n).

@ Dinesh Shrestha 15
Variance and standard error of the sample
proportion (for single proportion):
 Let 𝑥1, 𝑥2, … , 𝑥𝑛 be the random samples of size ‘n’
drawn from the population with proportion (P).
 Variance of the sample proportion in case of
without replacement is given by
 V(p) = PQ  N − n  if P is given.
n  N −1 

 Standard error of the sample proportion is given by

 S.E. (p) = PQ  N − n 
.  
n  N −1 

@ Dinesh Shrestha 16
 If N is not given/known or infinite or sampling is
done with replacement then
𝑃𝑄
 V(p) =
𝑛

𝑃𝑄
 S.E.(p) =
𝑛
 Where,
 n = sample size N = population size
 P = Population proportion.
 Q = 1- P

@ Dinesh Shrestha 17
 When population proportion (P) is not known then the
standard error of proportion is estimated by using
sample proportion (p).

 S.E. (p) = pq  N − n  if N is known/finite.

 
n −1  N 

𝑝𝑞
 S.E. (p) = if N is unknown/infinite.
𝑛−1
 Where,
𝑥
 p=
𝑛
 𝑥 = no. of certain characteristics in the sample.
 q=1-p

@ Dinesh Shrestha 18
Standard error for double proportions:
 If 𝑝1 and 𝑝2 be the sample proportion of sizes 𝑛1 and
𝑛2 then the standard error of difference between two
sample proportion is given by
 S.E.(𝑝1 - 𝑝2 ) = PQ
1 1 PQ
+ 2 2 if 𝑃1 and 𝑃2 are given/known.
n1 n2

1 1 
 = PQ  +  if 𝑃1 = 𝑃2 = P
 n1 n2 

1 1
 S.E.(𝑝1 - 𝑝2 ) = 𝑃෠ 𝑄෠ 𝑛1
+
𝑛2
if 𝑃1 and 𝑃2 are not given/unknown.
 Where,
𝑛1 𝑝1 +𝑛2 𝑝2 𝑥1 +𝑥2
෠
 𝑃= =
𝑛1 +𝑛2 𝑛1 +𝑛2
෠ = 1 - 𝑃෠
 𝑄

@ Dinesh Shrestha 19