GLS for cross-sectional data

[FatimaZohra SABRANE]

Dear Statalist

In a cross-sectional data and after using 'reg' I have found proof of heteroscedasticity (using 'hettest' command). I used robust standard errors to adjust for heteroscedasticity but I am still not confident about the results.
I read that GLS is more efficient than OLS under heteroscedasticity but it is only allowed for panel data.
Is there any other command that would correct for heteroscedasticity in the results?
I was thinking of 'hetreg' but not sure about it either.

Any guidance would be appreciated.

[Andrew Musau]

If you can pin-point exactly what is causing the heteroskedasticity, you can do something about it, e.g., redefining variables or using weighted regression. For the latter, see https://stats.oarc.ucla.edu/stata/ad...es-regression/. If your model is correctly specified, there is no reason to distrust heteroskedasticity-corrected standard errors.