sixtystrat wrote:As I referred to in an earlier post, I am having trouble using anything other than a PIM matrix in the simulation routine. Even when I use the Burnham model as suggested the same thing happens. Any ideas? I need to use a trend model which is not possible with a pim matrix. THanks!
Joe
My guess is you're forgetting that if you use a DM, you'll need to specify the beta values on the logit scale, not using the default identity link. If you try simulating under a DM with beta's entered correctly, and specified, on the logit scale, works fine. If you try it on the identity scale, crash and burn. If you think about it for more than 30 seconds, you should understand why.