www.cmm.bristol.ac.uk/forum

Hello. I am trying to compare 3 models in order to establish whether or not to pursue two multi-level structures. The models are as follows:
1: a null single level model using the regress command
2: a null two-level model using the runmlwin command (school and student)
3: a null cross-classified model (with two sets of level 2 units) using runmlwin mcmc (school, neighbourhood, student)

Can anyone suggest the best way to do this please? I have been comparing my simple two-level models (with predictor variables added) using lrtest and have been comparing my cross-classified mcmc models using the DIC, but I am unsure how to decide if I should pursue the cross-classified model versus the two-level model in particular.

The VPC does suggest a very small amount of neighbourhood clustering in the cross-classified model (<1%), but I can't decide if this is substantial enough to pursue the cross-classified model.

Thanks in advance for your help.

Best wishes,
Marie

Hi Marie,
To compare models 2 and 3 you can fit them both using runmlwin and MCMC and then compare the 2 DIC values.
Regards,
Bill.

thanks Bill, much appreciated!