Which model to choose? First Diff, Fixed Effects, Random Effects

[Pratap Pundir]

When doing panel analysis, is there a good rule of thumb to determine whether to use first differences, fixed effects or random effects, such as based on how balanced the data are etc?

[Carlo Lazzaro]

Pratap:
1) Stata can manage both balanced and unbalanced panel datasets without any problem;
2) the usual comparison is between -fe- and -re- specifricationn; -hausman- (or, if you invoked non-default standard errors, the community-contributed programme -xtoverid-) can help you out to choose between them.