Marginal effects xtlogit

[Helen Subbotina]

Hello, can I use mfx function after regression xtlogit? Is there other function to interpret marginal effects on panel data?

Thank you very much in advance!
Helen

[Clyde Schechter]

Assuming you are using Stata 12 or later, you are probably better off with the -margins- command. -mfx- is a user-written command which mainly provided some of the functionality of -margins- before Stata Corp. wrote the latter.

[Helen Subbotina]

Thank you very much, Clyde. And what will the interpretation of results be?
It seems this command provides me margins data for constant term only.

[Helen Subbotina]

Thank you very much, Clyde. And what will the interpretation of results be?
It seems this command provides me margins data for constant term only.

[Clyde Schechter]

-margins- has the ability to provide pretty much any kind of marginal effect you can dream up. You just need to specify it properly in the command. If you just do -margins- with no varlist and no options, you will get, as you noted, margins for the constant term only. Do read the help for margins, and, some of the examples in the manual. It will take a little time to learn to use this command, as it can get complicated. But it is well worth the effort. After doing that reading, try to get the specific marginal effects you are looking for. If you can't get it to work, then post again showing the commands you tried, what Stata gave you in response, and what you were hoping to get.

By the way, another very helpful resource on the use of the -margins- command (the clearest, in my opinion) is https://www3.nd.edu/~rwilliam/stats/Margins01.pdf by Richard Williams.

[Helen Subbotina]

Thank you very much, Clyde. This is very useful.

[Richard Williams]

Clyde, thanks for the plug. A few minor points: mfx is actually Stata's own old command for marginal effects. If somebody is condemned to using a very old version of Stata, they should check out mfx, and also possibly the user-written commands margeff and mfx2. But otherwise use margins.

I haven't installed Stata 14 yet, but it looks like the help for margins has greatly improved. It used to be you got this one size fits all help for margins. Now there is customized help for each estimation command. That can probably be very helpful for things like the xt commands, where the defaults may differ from the non-xt counterparts. Stata 14 margins is also much better for use after commands like ologit and mlogit.