EFried posted on Thursday, July 12, 2012 - 1:20 pm
We want to run a multi level analysis with 5 time points, 9 dependent variables, and 7 covariates. The dependent variables are ordered (0,1,2,3).
Numerical Integration isn't able to handle this amount of integration points (over 10k), neither is Monte Carlo integration.
The next possible estimator is Bayes, which does converge in about 2 days. However, as far as I understand the way the Bayes estimator performs in (1) multivariate models and (2) with ordered variables isn't well understood.
Is that correct? Would you advise against using the estimator in this case? Or is there literature showing it can be used?
EFried posted on Thursday, July 12, 2012 - 3:40 pm
(Sorry, of course the Bayes estimator does not "converge", but it takes that amount of time for the iterations to run through)