Regression: fixed effects + quantiles

Specific Domains Statistics
17

I absolutely think we need a better way of using fixed effects in GLM and statsmodels mode generally. I think either fe(x) or categorical(x) or something should be used to refer to fixed effects. For reference, the way to do this in StatsModels currently is

julia> lm(@formula(income ~ edu + state), df, contrasts = Dict(:state => EffectsCoding()))

Also, you have a typo in your post

reg(df, f1, Vcov.cluster(:state))     #Error

This works fine, does not error.

I definitely think that

  1. StatsModels should support fe
  2. FixedEffectModels should support glm and the general ecosystem better.
2 Likes