Predicting standardized residuals after mlogit (multinomial logistic regression) using svyset?

[Maria Serrano]

Dear users, I kindly ask for your guidance regarding my following problem:

I am conducting a multinomial logistic regression using mlogit and since I have a complex sample, I am also using the svy command to adjust the standard errors. I would like to test the assumptions for the multinomial logistic regression and therefore I would like to check out the residuals for any extreme values. Do you have an idea how I could it? For logit I can use the command "predict" after the logistic regression, but this command does not work for mlogit or svy.

I have read that you can also check values of cooks distance or DF beta to find extreme values, but I do not know how to calculate them after mlogit and svy either.

I would really appreciate any advice!

Kind regards,
Maria

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. You don't even tell us exactly what code you're using to estimate your model.

As far as I can see, any "outlier" problem would have to be on the rhs in an mlogit. You can't have extreme residuals since they the dv's and predicted values are probabilities and thus bounded. You could run a regression and then use some of the rhs diagnostics to look at colinearity. This is clearly an ad hoc approach but might work.