Identify observations counted in e(N_full) result in ppmlhdfe command

[Maria Veihl]

I am using the ppmlhdfe (http://scorreia.com/help/ppmlhdfe.html) estimation and reporting the number of observations, including dropped, singleton and separated obs. e(N_full). I would like to know which observations are included in the count of e(N_full) number. How can I generate the variable that is equal to 1 if the observation is counted in N_full?

[Andrew Musau]

Just repeat the estimation, keeping singletons.

qui ppmlhdfe ..., keepsing
gen fullsample= e(sample)

[Maria Veihl]

Thank you for the response. Maybe I am missing something, but your approach just provides me with the observations used in the estimation, which is different than what e(N_full) count is…

[Andrew Musau]

We are talking about 2 different regressions here. \({\bf \text{e(N_full)}}\) in regression 1 (default) is equal to \({\bf \text{e(N)}}\) in regression 2 (with option -nosing-). If you cannot follow this, provide a data example of a regression where \({\bf \text{e(N_full)}}\neq {\bf \text{e(N)}}\).