Small sample correction clustered avaialble?
joonro opened this issue · comments
Hi, thanks for the great package that is easy-to-use.
I was trying both statsmodels and linearmodels.PanelOLS to estimate the same fixed effects model, to make sure I get the same results. I found that when I use clustered standard errors, linearmodels was producing smaller standard errors for the same model.
It seems the issue is that in statsmodels, use_correction option was used, which provides small sample correction. When I turn that option to False, I get very similar standard errors from both packages.
I looked at the documentation but couldn't find any mention of applying the small sample correction for linearmodels when clustered standard errors, I was wondering if this is not available?
Adding group_debias=True to .fit() may help, as reported in #540.
Please see #540. Closing as a duplicate.