First Midterm (Practice With Solution)
First Midterm (Practice With Solution)
Econometrics 410
Thursday, Oct. 7
(a) If Assumption MLR.5 (Homoskedasticity) does not hold, then the ordinary least
square estimators are biased.
[A:] False. The OLS estimators are unbiased under Assumptions MLR.1-4
(linearity, random sampling, no perfect multicollinearity and zero-conditional
mean). Homoskedasticity is not necessary for unbiasedness to hold.
(b) Because omitted variables cause bias, it is always recommended to include all
available explanatory variables in a regression.
[B:] First of all, omitted variables do not always cause bias. Omitting a variable
only causes bias when the omitted variable is correlated with the included vari-
ables. Second, there are reasons that some variables should not be included in
a regression. One reason is that some variable may be an intermediate variable
between the important explanatory variable and the dependent variable.
(c) R-squared always increases when we add variables to a regression.
[A:] Strictly speaking, it is true if "increases" is replaced with "does not de-
crease". R-squared is the proportion of variation in the dependent variable that
is explained by the model. When we add variables to regression, at least the
same amount of variations is explained as before.
And you would like to test the hypothesis that math score and writing score have
the same eect on log (wage). That is, you want to test H0 : 1 = 2. We discussed
a way to do this in STATA by running a modied regression.
1
[A:]
where lwage denotes the natural logarithm of wage, educ is education, exper is
working experience in years.
(c) (20 min) The following output for equation (1) was obtained in Stata, where
several entries were replaced with letters. Find A-G. For example:
If you dont have a calculator, show all but the nal calculating step. For
example, in the above equation, you get full credit if you show all steps but
"=0.041009".
2
Source | SS df MS Number of obs = 526
-------------+---------------------------------------F( 3, 522) = D
Model | 44.5393713 A 14.8464571 Prob > F = 0.0000
Residual | 103.79038 B .198832146 R-squared = E
-------------+--------------------------------------Adj R-squared = 0.2963
Total | C 525 .28253286 Root MSE = .44591
------------------------------------------------------------------------------
lwage | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
educ | .0903658 .007468 F 0.000 G .1050368
exper | H .0051965 7.89 0.000 .0308002 .0512175
exper2 | -.0007136 .0001158 -6.16 0.000 -.000941 -.0004861
_cons | .1279975 .1059323 1.21 0.227 -.0801085 .3361035
------------------------------------------------------------------------------
[A:]
A = #regressors = 3;
B = n A 1 = 526 3 1 = 522
C = SST = SSE + SSR = 44:5393713 + 103:79038 = 148:33
SSE=A 44:5393713=3
D = F = = = 74:668
SSR=B 103:79038=522
SSE 44:5393713
E = = = 0:30027
SST 148:33
^ :0903658
1
F = = = 12:1
se ^ :007468
1
Note that when computing G we use the critical value from the standard normal distribu-
tion. This is OK because n = 526 is large.