Entertain a sufficient representation of categorical variables

[laszlosandor]

FYI: https://arxiv.org/pdf/1908.09874.pdf

I have limited intuition about this issue. Probably some of the large standard errors I've seen in models with many-many DFs come from the noise in a large N, small T scenario. Even if there is no cookie-cutter solution to this problem (yet?), it is so easy to run into with reghdfe that maybe it is prudent to advertise some remedies that people tried.