Hi,
I am trying to run a negative binomial population-average model in Stata 14. However, I am not sure what is the best approach to do this. Based on the comments from here: https://www.stata.com/statalist/archive/2010-12/msg00918.html, we estimated alpha using nbreg and then used the value of alpha in the xtgee model as follows: "xtgee v1 i.visit Age, i(ptid) corr(indep) fam(nb `alpha') robust"
Is this correct? Or should "alpha" be replaced by the default value of "1" in the xtgee?
We also tried using xtnbreg (instead of nbreg, then xtgee) - and the output for this is identical to that of the nbreg/xtgee with a value of 1 (instead of "alpha" estimated from nbreg). This output does differ quite a bit from using the value of "alpha" in the xtgee. Can anyone advise? Thanks.