Open droodman opened 4 months ago
Thank you, I agree on all points. I should say however that the coming weeks are quite busy and I don't know whether I'll have time for FOSS development....
Understood. Perhaps it will inspire you when the time comes to try the latest reghdfejl
, which does call GLFixedEffectModels
. It can be installed with
net install reghdfejl, replace from(https://raw.github.com/droodman/reghdfejl/master)
Then you can do the third example in the ppmlhdfe
help file:
use "http://fmwww.bc.edu/RePEc/bocode/e/EXAMPLE_TRADE_FTA_DATA" if category=="TOTAL", clear
egen imp = group(isoimp)
egen exp = group(isoexp)
ppmlhdfe trade fta, a(imp#year exp#year imp#exp) cluster(imp#exp)
reghdfejl trade fta, a(imp#year exp#year imp#exp) cluster(imp#exp) family(pois)
Here the runtimes are trivial. But in an example with ~10,000,000 observations, I found that using your package in this way halved run time.
(I still need to document the new functionality and make the regression output more appropriate, e.g., by showing the log likelihood.)
Right now the other two ppmlhdfe
help file examples fail because of issues raised above.
I'm adding support for this package to
reghdfejl
for Stata(!).This leads me to a few comments;
nlreg
accepted the GLMoffset
argument.dof_fes
field with the number of degrees of freedom consumed by the absorbed FE? FixedEffectModels has this.Also, I think it shouldn't crash when passed a formula with no
fe()
terms.