Page 1 of 1

Multilevel negative binomial model selection using R2MlwiN

Posted: Thu Oct 15, 2020 2:38 pm
by FerhatTura
Hi,

I am new to R/RStuido/R2MlwiN and running two-level negative binomial models. The Coef. and Std. Err. values of the independent variables are identical with MLwiN results (including intercept of the fixed and random parts) but not sure how to decide which model fits the data better as no Log likelehood and Deviance statistic is reported (i.e. NA). Without deviance statistics, is there a way of deciding which model fits the data better? Any help would be much appreciated.

Kind regards,

Ferhat

Re: Multilevel negative binomial model selection using R2MlwiN

Posted: Mon Oct 19, 2020 12:43 pm
by ChrisCharlton
If you're running the model via MCMC then you could compare the DIC statistics, otherwise I think that your only option would be a Wald test.

Re: Multilevel negative binomial model selection using R2MlwiN

Posted: Mon Oct 19, 2020 3:37 pm
by FerhatTura
Thank you very much Chris.