What's wrong here? Margins after xtlogit produce different results by different users

[Meng Song]

Dear all,
I'm trying to get this sorted. It appeared to me and I'm sure to other people that we get the same results running "margins, dydx(*)" after xtlogit, which are not correct marginal effects, while other users get different results after doing the same thing. Here is a link that you can see the problem:
https://www.statalist.org/forums/for...ter-xtlogit-re
Unfortunately, I'm the one who is getting the exact same results after running "margins, dydx(*)" following xtlogit. Could anyone help how to solve this, please? Many thanks.
I understand that some stata staff suggested using predict(pu(0)) in the margins option, but this is not the same as "margins, dydx(*)".
Best wishes,
Meng Song

[Richard Williams]

I posted a response in the thread you linked to. Different versions of Stata have different default options for margins after xtlogit. If different people seem to get different results, check what the Expression line of the margins output says. For some, the expression may tell you it is for the linear prediction, while for others it may be the predicted probability. (If you had your life to live over, I would recommend just posting only in this thread or the other one, as it can be confusing when the same discussion is going on in two different places.)

[Meng Song]

Thank you very much Richard. I apologise for the confusion. I'll definitely post only in the relevant thread in the future. Many thanks.