andrea.wainwright@CAPITALONE.COM wrote back: >I'm trying to run a Bradley-Terry model. In the example in the book >they have only 8 items and I have way more, but otherwise we are running >very similar data.

You want a Bradley-Terry model?

It didn't seem like you were coding that correctly. You might want to look in Alan Agresti's book "Categorical Data Analysis" to make sure you're getting it right.

Yes, you end up having to set Y=1 identically. But you have to keep the counts as your FREQ variable.

Also, the 600 variables seems really unwieldy and unworkable. How do you get that many different variables? How do you expect to be able to interpret all of them?

I would recommend that you try doing a Bradley-Terry model using PROC GENMOD. Here's a simple form for when you have a number of trials and a number of successes for each set of trials.

proc genmod; model successes/trials = / dist=bin link=logit noint; run;

Note that this moves the count information into the response part, and it assumes all your other numeric variables are regressors. Then you don't have to spend a huge amount of time naming or re-naming them.

HTH, David -- David L. Cassell mathematical statistician Design Pathways 3115 NW Norwood Pl. Corvallis OR 97330

_________________________________________________________________ Express yourself instantly with MSN Messenger! Download today - it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/