Hi All,
I am trying to estimate a Random Coefficient Model in a Negative Binomial framework and the variance structure looks to be a little wonky. I have a few questions and am hoping for a little guidance to run this model.
1. After switching to Neg. Binomial Regression, designating Y and offset. Should B0 be set as a constant? Does MLWin add a constant/and or error term into the equation for you? I guess the question is, do I need to add anything else to the model besides my predictors to get the model to run.
2. After allowing one of my level-one variables to vary (by selecting jState), when I add level-2 variables they are still predicting a random intercept correct?
3. Is it possible in MLWin to add the cross-level interactions in order to specify a "Slopes as Outcomes" model and examine how the level-2 variables predict the random slope/coefficient of the level one predictor.
Thanks for any help on this topic. I appreciate it, I am getting some results, but I want to be sure that I am doing what I think I am.
Cheers
Kevin