zero-truncated mixed effects logistic regression?
Martin Schmettow <schmettow at ...> writes:
Thank you for your ideas. You seem to suggest that I use a C-R model to estimate the number of missing values per category and then impute these values? Sounds feasible, although I was in hope for a more straight forward way to deal with zero trunc. binomial data like seemingly is available for Poisson data. Since zero trunc. Poisson is available in glmmADMB, would it also be an option to treat binomial data as Poisson by using a per class offset? CU, Martin.
Seems reasonable. I would try it on some simulations ... glmmADMB has (in principle) zero-inflated binomial (although I've never tested it ...), and truncated binomial could be implemented without a great deal of trouble ...