I am fitting a spatio-temporal Bayesian Poisson model with 22 explanatory variables, an offset variable, 2200 observations and non-informative priors. I am using the package CARBayesST which uses MCMC. I tried different specification for the spatiotemporal effects. With the exception of one of the specifications, I have not been able to achieve convergence. I check that by looking at Geweke's diagnostics and by looking at the trace plots of the coefficients. An example of one of the trace plots follows. The trace plot for three of the variables I am doing 4000000 sample size, 800000 burn-in, and thin = 10. I started from 500000 with 100000 burn-in and thin = 10. I have been increasing to 4 million but it doesn't seem it is helping with convergence. Should I keep increasing the sample size, or based on the trace plots (they are very similar to the one attached for all coefficients), I should change thin size, burn-in? Is there anything else I need to do?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.