PROC ROBUSTREG ==> can I get some kind of prediction limit or similar?

[iiibbb]

Hi--

I am new to the ROBUSTREG procedure... which has been very helpful addressing leverage points and outliers that are real and I can't simply disqualify them from my data.

Is there a functional equivalent to PROC REG's ucl lcl uclm lclm outputs for ROBUSTREG or is it irrelevant to the weighted least squares approach?

Alternatively I was just going to go with Quantile Regression to capture the variability.

[PaigeMiller · Accepted solution]

PROC ROBUSTREG provides estimates of the standard errors of the predicted mean (or standard error of predicted individual values) in the OUTPUT statement (STDM or STDI options). From those you can compute whatever confidence interval or prediction interval you want.