Submitted by: Submitted by ellenhua
Views: 10
Words: 2627
Pages: 11
Category: Business and Industry
Date Submitted: 11/25/2015 04:23 PM
Statistics and Econometrics (ECO00037)
Lecture 10: Simple Regression Analysis:
Inference
Lecturer: Takashi Yamagata (Room A/EC/018)
E-mail: takashi.yamagata@york.ac.uk
O¢ ce Hour: Tuesday 9.30-11.30
Reading: Dougherty Ch2
Autumn 2015
1 / 24
Review: Assumptions
yi = β1 + β2 xi + ui
I
I
I
I
I
I
I
A.1 The model (above) is correctly speci…ed.
A.2 There is some variation in the regressor in the
sample
A.3 E (ui ) = 0 for all i.
A.4 The disturbance term is homoscedastic: E (ui2 ) = σ2
for all i
A.5 ui is independent of uj for all i 6= j.
A.6 (For small sample) ui is normally distributed.
2 / 24
Sampling Distribution of the OLS Estimator
I
Since xi are nonstochastic, Assumptions A.1-A.6 imply
that
yi
i.i.d.N β1 + β2 xi , σ2 .
I
In addition, we can prove that, under Assumptions
A.1-A.6 we have
b2
therefore,
where
N [ β2 , Var (b2 )] ,
b2 β2
p
Var (b2 )
Var (b2 ) =
N (0, 1),
σ2
∑N 1 (xi
i=
x )2
¯
(1)
.
3 / 24
Sampling Distribution of Standardised Estimator
I
Why? Recall that
N
b2 = β2 + ∑i =1 ai ui with ai =
I
I
¯
(xi x )
.
¯
(xj x )2
∑N 1
j=
Since b2 is a linear function of ui
i.i.d.N (0, σ2 ), b2
should be normally distributed.
In practice the results (1) may not be very useful since
Var (b2 ) is usually not observable.
When we replace Var (b2 ) with its estimator,
d
ˆ
Var (b2 ) = σ2 / ∑N 1 (xi x )2 , we have the following
¯
i=
result:
b2 β2
q
tN 2 .
(2)
d (b2 )
Var
4 / 24
Sampling Distribution of Standardised Estimator
I
N denotes the sample size, and 2 is the number of
parameters to estimate. N 2 comes from the
N
I
e2
i
ˆ
denominator of the estimator of σ2 = ∑N=1 2 i .
As discussed in the previous lecture, we call the square
root of population variance ‘
standard deviation (s.d.)’
,
and estimator of the latter ‘
standard error (s.e.)’ Then,
.
we can rewrite (1) and (2) as
b2 β2
sd (b2 )
N (0,...