🚀 go-pugleaf

I've looked around for S-PLUS code that estimates models (e.g. GLM) from 
survey sample data that are weighted. This is an old debate. There doesn't 
seem to be a way to do this. I'd like to plead with some of the hot shots 
and brainiacs on this list. You can do so much with S-PLUS, but you can't 
even handle a simple random survey with different selection probabilities ? 
My data are from a simple random sample postal survey where a simple random 
sample of the initial postal nonresponders were selected for telephone 
followup.

I've done a lot of work in S-PLUS, but have to convert Stata to estimate 
the final models. S-PLUS is elegant, the code is relatively low 
maintenance, its graphs are production quality, you can interface with 
Latex and PDF. If you want, you can estimate more advanced models including 
overdispersion and GAMs. Stata has neither of these. Stata is fine if you 
have simple models. But including three-way interactions or splines makes 
Stata walk through uncharted territory, and it is just not elegant.

As a first step, I'd like to propose a group that will add to S-PLUS the 
possibility of doing a GLM with with correct variance estimates that has a) 
unequal selection probabilities (fractional weighting) and/or b) 
clustering. If somebody tell me what to do, and they can look over my 
shoulder, I will write it.

Some have said that all the survey people use SAS anyway. Well, that may be 
true if you just want GLMs and means and proportions. But shouldn't we 
bring the survey sample people into the S-PLUS fold? There is lots more to 
statistics than this. I'm sure that the leading edge of survey sampling 
people will take up on S-PLUS and use its powers. Then this will "trickle 
down" to the grunts (like me) in time.

Yours,
Jan Brogger, University of Bergen, Norway
---------------------------------------------------------------------
This message was distributed by s-news@lists.biostat.wustl.edu.  To
unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with
the BODY of the message:  unsubscribe s-news
---