What are "robust" standard errors?

[nilshg]

I had sort of assumed that they are HC3, but comparing to R::sandwich I get different (in my case slightly lower) standard errors from Vcov.robust() than sandwich's vcovHC()

[matthieugomez]

It should be the same as Stata. I'm not sure about R but they maybe there are differences in terms of small sample adjustment etc. Would be better to give the exact definition / formula in the documentation though, I agree.