Went back and did a little reading on the poisson model--sad to say, my
model (below) doesn't pass the post-estimation tests. So I tried a
negative binomial model (thanks to Ying Lu for teaching us all this
stuff), but I get two different sets of results, depending on how I set
the dispersion. Think I'll need to wait to get back home to reread my
notes on this stuff.
1 comment:
Glenn: what is your dependent variable? I'm sure its back further somewhere but I don't have the wherewithal to go through them all. I believe, off the tip top of my head, that negative binomial and poisson models are intended for overdispersion, not skewness. Have you simply tried transforming your dependent variable using the natural log?
Post a Comment