Marginal effects after fixed effect logistic regression

[Tapas Paul]

I was trying to run fixed effects logit with stata's

clogit

Now, trying to figure out the marginal effects (average partial effects). Thanks in advance.

[Phil Bromiley]

Read the documentation for clogit postestimation and for margins.

[Tapas Paul]

clogit y ib1.x1 i.x2 x3 x4 x5 if x1==1, group(id) nolog
margins, dydx(*) post

Stata output is telling that the average marginal effects are assuming fixed effect is 0 [Expression : Pr(x1|fixed effect is 0), predict(pu0)]. I'm defining the marginal effects using the fixed effect logit.

[Richard Williams]

Steve Samuels explains why marginal effects after clogit is not straightforward:

https://www.statalist.org/forums/for...-after-xtlogit

[Joao Santos Silva]

Dear Tapas Paul,

You may be interested in the aextlogit command and on this presentation:

https://www.stata.com/meeting/uk16/s...antos_uk16.pdf

The upshot is that in general it is not possible to compute marginal effects after the FE logit because the marginal effects depend on the fixed effects that cannot be estimated.

Best wishes,

Joao

[Tapas Paul]

Dear @Joao Santos Silva, The presentation helps. Can you tell me how things will be changed for survey data where I have survey weights? I run the survey weight first and then run aextlogit. Do you think this is the right approach? However, aextlogit doesn't work with svy (like, svy: aextlogit) Thanks.

Paul

[Joao Santos Silva]

Dear Tapas Paul,

I believe you would have to adapt the code to work with weights.

Joao

[Ah Khozari]

Dear @Joao Santos Silva,

i am trying to run the aextlogit command, but i get the error (command aextlogit is unrecognized). Is there any Package that i should install for the command?

thanks

[Joao Santos Silva]

Please install aextlogit by typing "ssc install aextlogit, replace"

[Ah Khozari]

Thank you Joao