In CVS (but not yet in the GUI) we have a new command "duration"
for estimating parametric duration models via ML. The default
distribution is Weibull, but there are also --exponential,
--loglogistic and --lognormal options. Right-censoring is
supported. I've tested some examples against stata and get good
I'm currently working on trying to get $yhat and $uhat to produce
something sensible. Suggestions would be gratefully received. I'm
having trouble grokking the discussions of generalized residuals
for duration models that I've come across, and I can't make much
sense of what stata produces with its "predict" function.
Show replies by date