melogit vs meqrlogit

[Dario Maimone Ansaldo Patti]

Dear All,

Due to some problem of convergence with melogit, i am thinking of using meqrlogit instead. I was able to compare the results of an estimation run with both of them and the point estimates are identical. However the standard errors are not. With melogit i can use clustered standard errors, while it seems to me that there is not the possibility of using robust standard errors with meqrlogit. Am i correct?

If I use meqrlogit, is the following statistical inference correct? I am afraid that without applying robust standard errors, my inference could be questionable.

thanks for your suggestions.

Dario

[Andrew Musau]

I believe that meqrlogit was absorbed into melogit via the option -intmethod(pcag)-, so there should be no need to use meqrlogit in the first place. On using a robust variance estimator, unless you have clustered data or suspect that your model is misspecified, you probably should not. See https://www.stata.com/support/faqs/s...nce-estimator/ for more on this.

[Dario Maimone Ansaldo Patti]

Andrew Musau thanks a lot for your explanation