e(sample) not working for ivreg2hdfe

[James Park]

For ivreghdfe, I could find summary stats of the regressor by running something like this

ivreghdfe Y (X=Z) C, absorb(FE1 FE2) cluster(Cluster)
summarize X if e(sample), detail

But for ivreg2hdfe, it seems it doesn't work.

ivreg2hdfe, depvar(Y) en(X) ex(C) iv(Z) id1(FE1) id2(FE2) cluster(Cluster)
summarize X if e(sample), detail

It seems e(sample) simply didn't work for ivreg2hdfe.

What can I do to obtain summary stats of X for the observations that were used in the regression that used ivreg2hdfe?


Random data to run this can be created by running the following.

clear

* Set a random seed for reproducibility
set seed 12345

* Generate random variables
set obs 100 // Number of observations
generate Y = rnormal(0,1)
generate X = rnormal(0,1) // Independent variable
generate Z = rnormal(0,1) // Instrumental variable
generate C = rnormal(0,1) // Control variable

* Create categorical fixed effects variables
gen FE1 = floor((uniform() * 5) + 1) // Categorical fixed effect with 5 categories
gen FE2 = floor((uniform() * 3) + 1) // Categorical fixed effect with 3 categories

* Create a cluster variable
gen Cluster = floor((uniform() * 5) + 1)

* Regression
ivreghdfe Y (X=Z) C, absorb(FE1 FE2) cluster(Cluster)
summarize X if e(sample), detail

ivreg2hdfe, depvar(Y) en(X) ex(C) iv(Z) id1(FE1) id2(FE2) cluster(Cluster)
summarize X if e(sample), detail

[Andrew Musau]

I do not use ivreg2hdfe from SSC as you are asked to explain in FAQ Advice #12, but what is the preference here over ivreghdfe also from SSC?

[James Park]

I do not use ivreg2hdfe from SSC as you are asked to explain in FAQ Advice #12, but what is the preference here over ivreghdfe also from SSC?

Yes they are both from SSC. They seem to give different coef and se...

[Andrew Musau]

ivreghdfe is essentially ivreg2 (from SSC) with reghdfe handling the absorbed fixed effects. The estimator has been sufficiently tested and works well. I do not know much about ivreg2hdfe, but you can use the official xtivreg including one set of fixed effects as dummies to compare the results.