Hello,

I'm running meglm models because it involves two levels (team and individual) and the dependent variable follows a gamma distribution.
However, STATA says it "cannot compute an improvement -- discontinuous region encountered."
I tried to remove one variable at a time to see what causes the problem and found that the team size variable, which is theoretically and empirically important, may be the source.
It shows results when I run the exactly the same models with Bayesian inference, but not with frequentist stats.
How can I fix this to see the result? Please advise.