Fixed effects-only regressions

[dzfinance]

I'm wondering if it is possible to conduct fixed effects-only regressions using this package without an explicitly RHS variable. The purposes of these analyses are mainly to understand the variations of the data (i.e., how many variations can firm FE explain, etc.).

Thanks

[s3alfisc]

Hi @dzfinance , I think this is already possible - you can get demeaned residuals (but not more, e.g. R2 values):

import pyfixest as pf 
import seaborn as sns

data = pf.get_data()

fit = pf.feols("Y ~1 | f1 ", data = data)
u = fit.resid()
y = fit._data[fit._depvar]

import seaborn as sns
sns.scatterplot(x = u, y = y)

What is the exact use case you have in mind?

[s3alfisc]

I think it should also be possible to call the predict method on an empty model.