Model Verifications GLS vs. OLS

[Simon Frei]

Hello everyone!

I'm doing an analysis of the impact of board composition (different types of diversity, etc) on company performance. As the data is in a panel structure (80 companies over 6 years) I intended to use xtreg.

Unfortunately, I am not very familiar with the generalized least square regression. I assume that as with every other regression type, you have to verify the model. Meaning check for: Multicollinearity, Autocorrelation, Non-linearity and Heteroscedasticity.

Reading through a number of studies, I learnt that for multicollinearity one uses the OLS model for VIFs and uses a correlation matrix as usual. Autocorrelation should be counteracted by the GLS regression, I assume. I did not find anything on non-linearity or heteroscedasticity and thus wanted to ask, if I need to check for these two criteria and how I can do so in Stata.

Thankful for any advice on how I could tackle this issue.

Best regards,
Simon

[Carlo Lazzaro]

Simon:
if you have a panel dataset with N>T, rarely pooled OLS outperforms -xtreg-. Under -xtreg- heteroskedasticity and/or autocorrelation can be dealth with robust/cluster standard errors.
That said (and referring to OLS outside the panel framework)
1) after OLS, heteroskedasticity can be tested via -estat hettest- (however, a visula inspection of residual distribution is recommended);
2) after OLS, non-linearity can be tested via -estat ovtest- (however, a visula inspection of residual distribution is recommended).