DFBETA problem

[George S]

I'm trying to solve this problem for another person so be gentle if there are big statistical issues involved please, but why does dfbeta() not work on regression results that were calculated with the "robust" option?

I get the error "option dfbeta() not allowed after robust estimation"

Thanks!

[Richard Williams]

Usually things like that happen when the assumptions behind a measure are violated.

It probably wouldn't be that horrible if you reran the regression without robust, especially if the robust standard errors aren't that different from the regular standard errors. It is just a diagnostic device anyway. What you do with the outliers once you find them is what is really what is important. If you are just planning to delete all cases that exceed a certain theshold then you've got a problem, but if you are going to examine the cases and figure out what, if anything, to do, then it isn't like your diagnostic measures have to be accurate to 12 decimal places.