www.cmm.bristol.ac.uk/forum

Hi there,

I'm using R2Mlwin to fit multilevel models using a binomial distribution. However neither the log-likelihood or deviance statistics are produced using this method. Are there any alternative statistics it is possible to extract from the model as an indicator of fit?

many thanks

A Wald test will often be used for these kind of model (see chapter 9 of the MLwiN manual examples on https://www.bristol.ac.uk/cmm/software/ ... /examples/).

As MLwiN fits discrete models in (R)IGLS using quasi-likelihood it is usually advised to fit them using MCMC, at least for the final models. In this case you would have the DIC statistic available to use as an indicator for model fit and comparison.

Many thanks Chris

best,

Tim