bias correction for one-way FE?

[swwilcox]

Enjoying alpaca, thank you so much for putting it together!

One small observation, possible request...

It seems that feglm() permits estimation with one-way fixed effects (thanks!), however,
I am encountering problems with getting the bias correction (using biasCorr()) with one-way fixed effects.

In particular, it seems the default for biasCorr() with the panel.structure argument is "classic", which expects a two-way specification. When I try to feed an feglm() object that specifies a one-way fixed effect (e.g. a location fixed effect) I get the following message:

"Error: panel.structure == 'classic' expects a two-way fixed effects model."

I am thinking in particular about settings that I am facing where the variation available is very sensitive to time fixed effects (wipes out too much variation).

In these kinds of settings, where for whatever reason, there is a good basis for only including either between or within fixed effects, it would be helpful to be able to recover bias-corrected estimates.

Thanks!

SW

[amrei-stammann]

Hi,

so far alpaca only permits bias-corrections for two- and three-way fixed effects models. We plan to add one-way in the future.

If you are interested in a one-way bias-correction, please check out our packge bife:
https://cran.r-project.org/web/packages/bife/index.html

Best wishes,
Amrei