PANEL DATA

[EViews Gareth]

It isn't clear to me what he means by "the t-stat is too high".

[tchaithonov]

Well, if you do a ls with stacked data (not pooled ols, just ls), then you t-stat would be too high for sure.

Dependent Variable: RI
Method: Panel Least Squares
Date: 03/22/10 Time: 12:55
Sample: 1/06/2003 1/03/2010
Periods included: 365
Cross-sections included: 59
Total panel (balanced) observations: 21535
White diagonal standard errors & covariance (no d.f. correction)
RI = RF + C(1)*(RM_RF)

Coefficient Std. Error t-Statistic Prob.

C(1) 1.006495 0.008404 119.7677 0.0000

R-squared 0.252317 Mean dependent var 0.173658
Adjusted R-squared 0.252317 S.D. dependent var 5.622736
S.E. of regression 4.861903 Akaike info criterion 6.000784
Sum squared resid 509023.0 Schwarz criterion 6.001154
Log likelihood -64612.44 Hannan-Quinn criter. 6.000904
Durbin-Watson stat 2.111100

The t-stat would be 119.77.

[EViews Gareth]

You can't really say whether the t-stat would be too high unless you know the underlying assumptions he is making about his model. Also, I don't believe you can just say "119 is too high".

We're venturing into the realms of econometric discussion rather than EViews discussion though.

[tchaithonov]

I agree. I guess it depends, as you said, on the underlying assumptions.

[ec07ssn]

Thank you so much for your help. Much appreciated! :)