Ordered logistic regression with robust standard errors?

[Andreas Head]

Hello,
I wonder why one should apply robust standard errors in ordered logistic regression models?
As far as I understood, there is no assumption stating that the residuals have to be homoskedastic.
Still, Stata offers the -robust- option for ologit models which made me wondering if I am wrong about homoskedasticity?

Any opinion you would like to share with me on that matter?

Thank you,
Andreas

[Richard Williams]

Here is one blog post that argues you should NOT use robust: http://davegiles.blogspot.com/2013/0...nonlinear.html

If hetero really is a problem, you may want to use programs like hetprob or the user written oglm.

[Andreas Head]

Tanks Richard, this is a great blog!