Open adamoppenheimer opened 1 year ago
Another set of options to add for Q matrices:
cov(alpha_i, alpha_j)
for coworkers, to estimate sorting of workers within firmscov(psi_j1, psi_j2)
for firms that are directly linked by workers, to estimate some sort of network effect of whether firms with similar firm effects are connectedAdd option to break down Q by each group of a particular category.
E.g. you could specify compute var(psi)
but separately by firm quantile, where you don’t include firm quantile as a control variable, but just use it for separating the Q matrix.
These look like useful features. I think it'd also be nice to compute variance components within clusters, such as regions.
@santiagohermo what is the difference between your suggestion and the last suggestion I wrote?
Also, the last suggestion I wrote is available in the dev version. If you check here, you can see how to use the updated Q code so that you can get the variance components within a particular category/cluster.
Nice @adamoppenheimer, thanks! The package keeps getting better. I'll take a look at the dev version.
Add option to estimate within and between firm variance for AKM and its bias corrections.