Professional Documents
Culture Documents
Econometrics
REGRESSION ANALYSIS
Estimate the population regression function (PRF) on the basis of the
sample regression function (SRF) as accurately as possible
The two-variable PRF:
Yi = 1 + 2Xi + ui
The PRF is not directly observable. We estimate it from the SRF:
Yi = 1 + 2Xi + ui = i + ui
Here i is the estimated (conditional mean) value of Yi
Choose the SRF in such a way that the sum of squares of the residuals, 2 = (Yi i) is as small as
possible.
Differentiating 2 partially with respect to 1and 2, we obtain
LPM
The linear probability model (LPM) parameters are estimated based
on OLS method
Run the OLS regression and obtain i= estimate of the true E(Yi | Xi).
Then obtain i=i (1 i), the estimate of wi.
Use the estimated wi to transform the data as shown in equation
below and estimate the transformed equation by OLS (i.e., weighted
least squares).
LOGIT
For each sample X, compute the probability of occurance as i=ni/Ni
where Ni is the sample space and ni is the number of occurrences.
For each Xi , obtain the logit as i= i /(1-i)
Then obtain i= Ni i /(1 i), the estimate of wi
Use the estimated wi to transform the data as shown in equation
below and estimate the transformed equation by OLS (i.e., weighted
least squares).
PROBIT
The estimating model that emerges from the normal CDF (cumulative
distribution function) is popularly known as the probit model,
although sometimes it is also known as the normit model.
We express the utility index Ii as a function of explanatory variables Xi
as Ii = 1 + 2Xi and the CDF for Ii is
10
11
After
Change
Group 1
(Treat)
Yt1
Yt2
Yt
= Yt2-Yt1
Group 2
(Control)
Yc1
Yc2
Yc
=Yc2-Yc1
Difference
Difference
Y
Yt Yc
12
Treatment effect=
(Yt2-Yt1) (Yc2-Yc1)
Yc1
Yt1
Yc2
Yt2
control
treatment
t1
t2
time
13
14
After
Change
Group 1
(Treat)
0 + 1
0+ 1+ 2+ 3 Yt
= 2 + 3
Group 2
(Control)
0 + 2
Difference
Difference
Yc
= 2
Y = 3
15
Y1 0 1 1
Y0 0 0
Thus the OLS estimate is
Y1 Y0= 1 + 1 0
This implies that standard methods (OLS) give an unbiased estimate of 1, which is the
average treatment effect
That is, the treatment-control mean difference is an unbiased estimate of 1,
Population
Sample
E[ ] 0
cov[ X , ] 0
1
T
1
T
'
t t
y X , so that y X
Moment conditions:
T
E[ ] 0 T1 ( yt X t ) 0
t 1
E[ X ' ] 0 X ' ( y X ) 0
MM estimator