Inflated t-statistics in pooled regression
Posted: Tue Mar 18, 2014 12:10 pm
Good Afternoon,
I had a question regarding my pooled regression results. It seems like all my variables are highly significant and I don't know whether I should be concerned about it. I'm thinking that there's a possibility that the standard error for each coefficient could be underestimated. I did check for correlation between explanatory variables and made sure to exclude variables that were correlated with each other. Is there any other test that I can perform to make sure that the t-stats are accurate? My sample only has 50 observations, however it has 800 total pool observations. Attached is the regression output. Thanks so much.
Vishal
-----------
Dependent Variable: PREMIUM1
Method: Pooled EGLS (Cross-section random effects)
Date: 03/16/14 Time: 23:51
Sample: 1 62
Included observations: 50
Cross-sections included: 16
Total pool (balanced) observations: 800
Swamy and Arora estimator of component variances
Variable Coefficient Std. Error t-Statistic Prob.
C -0.444429 0.061761 -7.195958 0.0000
DUMMY_PROPERTY 0.216286 0.028296 7.643769 0.0000
DUMMY_HR -0.333284 0.033430 -9.969465 0.0000
DUMMY_CR -0.183334 0.037304 -4.914613 0.0000
DUMMY_BRT -0.388335 0.080136 -4.845918 0.0000
DUMMY_SEMILOG 0.253909 0.043471 5.840823 0.0000
ACCESS 0.205326 0.032544 6.309235 0.0000
NEIGHBORHOOD_HHMI 0.011848 0.001303 9.091320 0.0000
DUMMY_DATE 0.108985 0.032383 3.365541 0.0008
Random Effects (Cross)
1--C 0.000000
2--C 0.000000
3--C 0.000000
4--C 0.000000
5--C 0.000000
6--C 0.000000
7--C 0.000000
8--C 0.000000
9--C 0.000000
10--C 0.000000
11--C 0.000000
12--C 0.000000
13--C 0.000000
14--C 0.000000
15--C 0.000000
16--C 0.000000
Effects Specification
S.D. Rho
Cross-section random 0.000000 0.0000
Idiosyncratic random 0.296835 1.0000
Weighted Statistics
R-squared 0.272507 Mean dependent var 0.171520
Adjusted R-squared 0.265149 S.D. dependent var 0.342972
S.E. of regression 0.294007 Sum squared resid 68.37426
F-statistic 37.03688 Durbin-Watson stat 2.026779
Prob(F-statistic) 0.000000
Unweighted Statistics
R-squared 0.272507 Mean dependent var 0.171520
Sum squared resid 68.37426 Durbin-Watson stat 2.026779
I had a question regarding my pooled regression results. It seems like all my variables are highly significant and I don't know whether I should be concerned about it. I'm thinking that there's a possibility that the standard error for each coefficient could be underestimated. I did check for correlation between explanatory variables and made sure to exclude variables that were correlated with each other. Is there any other test that I can perform to make sure that the t-stats are accurate? My sample only has 50 observations, however it has 800 total pool observations. Attached is the regression output. Thanks so much.
Vishal
-----------
Dependent Variable: PREMIUM1
Method: Pooled EGLS (Cross-section random effects)
Date: 03/16/14 Time: 23:51
Sample: 1 62
Included observations: 50
Cross-sections included: 16
Total pool (balanced) observations: 800
Swamy and Arora estimator of component variances
Variable Coefficient Std. Error t-Statistic Prob.
C -0.444429 0.061761 -7.195958 0.0000
DUMMY_PROPERTY 0.216286 0.028296 7.643769 0.0000
DUMMY_HR -0.333284 0.033430 -9.969465 0.0000
DUMMY_CR -0.183334 0.037304 -4.914613 0.0000
DUMMY_BRT -0.388335 0.080136 -4.845918 0.0000
DUMMY_SEMILOG 0.253909 0.043471 5.840823 0.0000
ACCESS 0.205326 0.032544 6.309235 0.0000
NEIGHBORHOOD_HHMI 0.011848 0.001303 9.091320 0.0000
DUMMY_DATE 0.108985 0.032383 3.365541 0.0008
Random Effects (Cross)
1--C 0.000000
2--C 0.000000
3--C 0.000000
4--C 0.000000
5--C 0.000000
6--C 0.000000
7--C 0.000000
8--C 0.000000
9--C 0.000000
10--C 0.000000
11--C 0.000000
12--C 0.000000
13--C 0.000000
14--C 0.000000
15--C 0.000000
16--C 0.000000
Effects Specification
S.D. Rho
Cross-section random 0.000000 0.0000
Idiosyncratic random 0.296835 1.0000
Weighted Statistics
R-squared 0.272507 Mean dependent var 0.171520
Adjusted R-squared 0.265149 S.D. dependent var 0.342972
S.E. of regression 0.294007 Sum squared resid 68.37426
F-statistic 37.03688 Durbin-Watson stat 2.026779
Prob(F-statistic) 0.000000
Unweighted Statistics
R-squared 0.272507 Mean dependent var 0.171520
Sum squared resid 68.37426 Durbin-Watson stat 2.026779