negative weights in logistic regression

[Sarah Roberts]

Hi everyone,

I need to conduct a series of regression analyses with negative weights. (I know this is unusual but it is conceptually sound.) The Stata manual says that negative weights are allowed when specified in the "svyset" function with the iweights option. (Syntax svyset id [iweights = w])

This worked with the svy:regress and svy:logistic commands, but when I tried svy:poisson or svy:glm, I got an error for "negative weights encountered". Is there a way to model count data with negative weights?

[Steve Samuels]

Welcome to Statalist, Sarah! This looks like a bug, and I've forwarded your post to Stata Tech Support ([email protected]).

[Rafal Raciborski (StataCorp)]

This is indeed a bug and we will have it fixed in the next update.

[Sarah Roberts]

Thanks very much, Steve and Rafal. Is there a way to find out when the next update might be released?

[Steve Samuels]

This problem was fixed in the May 19 Stata update