Different results after update (Feb 2016) for melogit and meologit estimates

[Carolin Rapp]

Hi,
I recently ran an update for Stata (Version 14.1). Afterwards I received completely different [different meaning opposite influence directions!] results for my estimates using meologit and melogit, respectively. Does anyone know what has been updated for these commands? I couldn't find any information on this issue. And had anyone, by chance, encountered the same problem?

Thanks for your help.
Best, C.

[William Lisowski]

Welcome to Statalist!

You'll find that the information displayed by help whatsnew will tell you the changes made in each of the releases since 14.0. It confirms that changes were made to melogit in the February 2016 release (as well as in several earlier releases, we don't know what release you updated from) but the description of the changes does not suggest they should produce changes in results of the sort you described.

Perhaps someone else will be able to confirm your experience. I need to ask, are you certain you are running the identical model with the identical data that you previously used? My experience has been that when I have a problem and I know something has changed (in this case, the Stata release), I'm disinclined to look for other changes that may explain the problem. I wish it were possible to roll back the update you installed, but Stata does not seem to provide a "downdate" capability. If you have a colleague who has not updated, or who is running an earlier version even, perhaps you could arrange for them to try your model on your data.