I'm trying to run some panel data, and the way the sampling occurred I should bootstrap the standard errors with resampling by cluster (as described in Harden 2011). I've done this successfully with xtreg (fixed effects) but I'm trying to incorporate instrumental variables and I'm getting some error messages.

The data I have is protected, but I can simulate the problem with the nlswork dataset.

Code:
cls  
webuse nlswork, clear  

keep if year==77 | year==78  

gen sampleclustervar=1 if idcode<2696
replace sampleclustervar=2 if idcode>2696 & idcode<6000  

xtset id year xtreg ln_wage age tenure hours union, fe vce(bootstrap, reps(1000))  cluster(sampleclustervar) // This line works fine
xtivreg ln_wage age tenure hours (union= south), fe vce(bootstrap, reps(1000))  cluster(sampleclustervar) // This line does not work

According to what I'm seeing in the help file, this should work, I think. But instead I get:

Code:
option cluster() not allowed
an error occurred when bootstrap executed xtivreg
And unfortunately it doesn't look like xtivreg2 supports it either, which is too bad because I really like the automatic tests that come with it. Does anyone have an idea of how to fix this? I'm using 15.1



(Note: I apologize for posting this twice. I didn't post the right tags on this earlier and now I can't seem to delete my original or edit it to put the tags on. The original was here: https://www.statalist.org/forums/for...ors-by-cluster)