Not really a SAS question, but I've gotten good advice here in the past

I am trying to model a variable that has a huge proportion of 0's, and a long right tail. I've tried fitting all sorts of models, in particular, zero inflated Poisson and zero inflated negative binomial models.

For all the models I've tried, the residuals are grossly nonnormal with HUGE values for both skewness and kurtosis. The problem is that there are a bunch of outliiers. I also tried fitting separate models for 0 vs 1 and then one for everyone with 1 or more. I also tried winsorizing the DV, making the highest value 50.

N is fairly large (at least, compared to what I'm used to) - it's about 5000.

The DV has a mean of 0.69, sd = 4.02, skew = 30.7, kurtosis =1294 (not a typo, kurtosis over 1000). Similar values hold for residuals from many of the models (OK, SOME models reduce kurtosis to 400 or so.....but you get the idea).

Any ideas?

TIA

Peter

Peter L. Flom, PhD Assistant Director, Statistics and Data Analysis Core Center for Drug Use and HIV Research National Development and Research Institutes 71 W. 23rd St www.peterflom.com New York, NY 10010 (212) 845-4485 (voice) (917) 438-0894 (fax)