Panel Dataset Regression with xtpoisson FE Problems with vce(robust)

[Finley Voigt]

Hello there,

I have a question, I am currently working on a regression within a Panel Dataset wich is set like this: xtset gvkey appyear.

The regression i want to do should be with the xtpoisson, I often heard that one has to cluster the standard errors through the notion: fe vce(robust), at the end of the regression.

My regression without the clustered SE looks like that:

But if I use the vce(robust) everything turns insignificant (P>z goes through the roof, Coefficient turn negative and vice versa etc.)

Do you have any tipps, do I even need the Robust function, do I need to use it differently? Help would be appreciated :D

Thank you very much!

[Carlo Lazzaro]

Finley:
you may want to take a look at https://www.statalist.org/forums/for...tandard-errors

[Joao Santos Silva]

Dear Finley Voigt,

In addition to Carlo's useful suggestion, I suspect that you are doing something wrong because you say that with vce(robust) coefficients change signs, which is not possible because that option only changes the standard errors, not the coefficient estimates. Please post exactly what you are typing and the results you get so that we can advise. In any case, I think you should use robust standard errors; there are only very rare exceptions to this rule when using Poisson regression.

Best wishes,

Joao