Hi,

I am not familiar with the terms that you use. But, in order to model for frequency of claims I would create a variable indicating whether the patient had a claim or not and use that as a dependent variable.

Hope this is useful.

-Premalatha

baogong jiang <bgjiang@GMAIL.COM> wrote: Hello:

I am trying to investigate the relationship between the drug compliance and ED visit. what I have now: 1. demographic informaiton on patients level (recip_id, age, sex ,race, and drug compliance) 2. claims records for ED visit (recip_id, er_visit (0/1) note: some patients have multiple ED visit;

My question: Can I aggregated the claims record by patients and then fit a negative binomial model? If I do not aggregate claims records, how can I fit a logistic model?

thank you the help,

baogong

--------------------------------- Yahoo! Messenger with Voice. Make PC-to-Phone Calls to the US (and 30+ countries) for 2¢/min or less.

--------------------------------- Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.