Page 1 of 1

DICs difference when comparing models after MCMC

Posted: Tue Mar 23, 2021 5:21 pm
by ManuelDewez
Hello,
I am comparing two models with a binary response variable with MCMC.
Model 1 is a single level model, model 2 is a two level models.
I understand that in this context, the DIC statistics should be used to compare models, and that the model with the lowest DIC is a better fit.

Is there a value for the difference in DICS that is used to consider that the difference between models is statistically significant (ie a value equivalent to p<0.05 in non-mcmc models)?

I know such "magical" values are somehow controversial, but they would be helpful to explain the results to readers that are not familiar with MCMC.

Thank you .

Manuel

Re: DICs difference when comparing models after MCMC

Posted: Tue Mar 23, 2021 5:58 pm
by ChrisCharlton
Bill gives some guidance on slide 57 of a workshop presentation that he gave some years ago, available at:

https://www.ukdataservice.ac.uk/media/3 ... ation4.pdf

The MRC Biostatistics unit, who originally developed the BUGS software, also give some suggestions:

https://www.mrc-bsu.cam.ac.uk/software/ ... ct-dic/#q9

Re: DICs difference when comparing models after MCMC

Posted: Wed Mar 24, 2021 9:22 am
by ManuelDewez
Thank you, very useful!