I am trying to run a JAGS model with two model comparison (both binomial dbern models with several categorical predictors). I am using the rjags package.
Could anyone suggest or give some references how to write such a model? I would greatly appreciate your input.
Dear frind
You just compare the accuracy of the two methods using R script.
i hope it helps you.
Best regards
Dr.Indrajit Mandal
Dear Dr.Mandal,
Thank you. I would like to run MCMC chain in Jags for both models at the same time to determine which model is more likely to be predicted for my data. I have looked at some examples from Krushke's Doing Bayesian with Jags and I am trying to modify it to categorical data.
The crucial question is what you mean by "compare". E.g. you don't even need data to compare the complexity of the models. If you want to assess which model fits your data at hand best, look at DIC as Mathieu suggested.
I suspect that you want to assess what the probability that one or the other model is generated the data. Bayes Factors are quite intuitive to assess this but note that there's a lot of different ways to calculate a lot of different bayes factors. One for each prior on your parameters at least.
The most explicit method (and therefore my favourite) is the product space method, where you have all models (two in this case) in the BUGS model and then a categorical indicator variable which selects either one. Then it's really a latent mixture model. You simply monitor the frequency each category being selected as you sample from the posterior. I've linked a helpful "tutorial" paper on the product space method below. Google it to find a free PDF.
http://www.sciencedirect.com/science/article/pii/S0022249611000423
As mentioned you can recourse to DIC, Posterior Predictive checks. Have a look at the Bugs book and Gelman's book "Bayesian Data Analysis"
http://www.mrc-bsu.cam.ac.uk/software/bugs/the-bugs-project-the-bugs-book/
http://www.stat.columbia.edu/~gelman/book/
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package