How to drop singleton observations?

[Alberto Alvarez]

Dear Stata Users,

I am using "reghdfe" packages that automatically drops singleton observations. I use the following regression:

reghdfe y x1 x2 x3 x4 , ab(firm year i.state#i.year i.industry#i.year) vce(cluster firm)

My question is, knowing the fixed effect structure from the above regression, how can I manually drop singleton observation in my data sample?

Thank you for the help.

[Carlo Lazzaro]

Alberto:
set aside any consideration about the correctness of your approach, you may want to try:

bysort firm (year): drop if _N==1

[Alberto Alvarez]

Thank you very much for your reply. I followed your step, by I still have some singleton observations that are dropped once I use "reghdfe" packages. Can it be because of "i.state#i.year i.industry#i.year"?

[Andrew Musau]

If you have missing entries on used variables but nonmissing identifiers, Carlo Lazzaro's code will not capture the singletons. A simple way is to determine the in-sample observations and then drop firms with a single observation. See

help e(sample)

A more complicated way is the following:

gen nonmissing= !missing(y) & !missing(x1) & !missing(x2) &  !missing(x3) & !missing(x4)
bys firm: egen count= total(nonmissing)
drop if count<2