# predict conditional mean after truncated negative binomial regression

I was using tnbreg command and wanted to predict conditional mean. As suggested by Stata for post-estimation after tnbreg code, I used predict condmen, cm to calculate the conditional mean, E(y|y > ll). When I manually calculated the conditional mean it was different than the output from Stata. So, I want to make sure that why the value I calculated is different than the Stata output. Could anyone please recommend me any supporting document describing the actual formula behind predict, cm?, then, I should be able to know that I am using the same formula manually as Stata does by its code.

• (1) It is Stata, not STATA. It isn't an abbreviation, and there is no need to yell. (2) Show your code: provide a reproducible example. (3) Migrate the question to CrossValidated (although this is up to moderators): there is zero content in it that relates to deal with open data sources per se. – StasK Apr 13 '17 at 13:33

The formula is given on p. 2388 of http://www.stata.com/manuals13/r.pdf. Without a reproducible example from you, it is impossible to discuss this further. You may have a bug, or they may have a bug -- looking at their formula, it seems like they've assumed that ll=0, as the formula quoted seems to assume the truncation point of zero.