Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 10

Classical Normal Linear

Regression Model (CNLRM)

Gujarati 4e, Chapter 4


The probability distribution of
disturbance ui

 The CNLRM differs from the CLRM in that


is specifically assumes that the
disturbance term ui entering the regression
model is normally distributed
 The CRLM: the mean value of ui is zero
and its variance is a finite constant

01/06/22 Prepared by Sri Yani Kusumastuti 2


The normality assumption for ui

 The CNRLM assumes that each ui is distributed


normally with
Mean: E  ui   0

 
E ui  E  ui    E ui2   2
2
Variance:

cov  u , u  : E  u  E  u    u  E  u     E  u u   0 i j
i j
 i i j
 j i j

 The assumptions given above can be more


compactly stated as

ut ~ N 0,  2

01/06/22 Prepared by Sri Yani Kusumastuti 3
Why the normality assumption?

1. Central Limit Theorem (CLT): if there are a


large number of independent and identically
distributed random variables, the distribution
of their sum tends to a normal distribution.
2. Even if the number of variables is not very
large or if these variables are not strictly
independent, their sum may still be normally
distributed.

01/06/22 Prepared by Sri Yani Kusumastuti 4


Why the normality assumption?

3. Any linear function of normally distributed


variables is itself normally distributed  if ui
are normally distributed, so are ˆ1 and ˆ2
4. The normal distribution is a comparability
simple distribution
5. If we are dealing with a small sample size
(less than 100 observations), the normally
assumption assumes a critical role.

01/06/22 Prepared by Sri Yani Kusumastuti 5


Properties of OLS estimators under
the normality assumption
1. They are unbiased
2. They have minimum variance
 They are minimum-variance unbiased or
efficient estimator
3. They have consistency  as the sample size
increase indefinitely, the estimators converge
to their true population values

01/06/22 Prepared by Sri Yani Kusumastuti 6


Properties of OLS estimators under
the normality assumption

4. ̂1 ˆ ~ N  ,  2ˆ
is normally distributed  1 1 1
 
Mean  
E ˆ1  1

Variance  2ˆ 
 X i
2

2
1
n x 2

̂ 2 is normally distributed  ˆ2 ~ N   2 ,  2ˆ 


i

5. 2

Mean  
E ˆ2   2
2
Variance  2ˆ 
2
i
x 2

01/06/22 Prepared by Sri Yani Kusumastuti 7


Properties of OLS estimators under
the normality assumption

6. 
 n  2  
ˆ 2 2
is distributed as the 2
distribution with (n-2)df
7.  
ˆ1 , ˆ2 are distributed independently of ˆ 2
8. ˆ1 and ˆ2 have minimum variance in the
entire class of unbiased estimators, whether
linier or not.

01/06/22 Prepared by Sri Yani Kusumastuti 8


Properties of OLS estimators under
the normality assumption

 To sum up: the normally assumption enables


us to derive the probability, or sampling,
distributions of ̂1 and ˆ2 (both normal) and
ˆ 2
(related to the chi square).
 With the assumption that ut ~ N 0,  2 , Yi  
being a linear function of ui, is itself normally
distributed

Yt ~ N 1   2 X i ,  2 
01/06/22 Prepared by Sri Yani Kusumastuti 9
Maximum Likelihood (ML)
 The probability distribution of the disturbance
term ui follow the normal distribution.
 Under the normality assumption, the ML and
OLS estimators of the intercept and slope
parameters of regression are identical.
However, the ML and OLS estimators of
variance of ui are different.
 The ML is called a large-sample method

01/06/22 Prepared by Sri Yani Kusumastuti 10

You might also like