bashtage
[Question] Why is AbsorbingLS in the iv package?
I use AbsorbingLS a lot for fixed effects analysis and I prefer it over PanelOLS because of its flexibility and performance.
But why is it located in the iv package? It's not used by IVxxx estimators, it doesn't provide an interface to absorb fes before IV2SLS or something like that, there is nothing instrument related in its API. It confuses me. What is the rationale?
And when is PanelOLS a better alternative than AbsorbingLS?
In my mind: fe analysis -> AbsorbingLS, iv analysis -> IV2SLS, that's how both questions are related.
Thank you in advance.
AbsorbingLS is inspired by areg in Stata, which is a way of running OLS models with high-dimensional dummies. It has post-estimation test statistics that are similar to OLS. While I agree that it is not exactly IV, it is also not really intended to be a panel data model since it doesn't have post-estimation of a panel model nor does it produce correct standard errors in general if the dummies are really fixed effects.
It perhaps should be on its own top-level subpackage.
PanelOLS is better when you have up to two fixed effects and you want to have correct inference.
Thank you for taking the time to answer.
nor does it produce correct standard errors in general if the dummies are really fixed effects
How come? Is it not correcting for the dfs used while absorbing the effects?
Or perhaps you refer to something like panel-corrected standard errors (https://www.stata.com/manuals13/xtxtpcse.pdf) ?