Dear everyone,
I would like to know whether it'd be possible to add iterations to a MCMC model that has already run through runmlwin. I know this is possible on mlwin but I was wondering if there was any way to do so with runmlwin.
So a practical example would be the following:
- I run a MCMC model for 100,000
- Upon inspection of the chains, I decide that convergence has not been reached
- I add another 100,000 iterations to reach a total of 200,000 (but without having to re-run my model from the start).
Thank you,
Léa
Adding iterations
-
GeorgeLeckie
- Site Admin
- Posts: 432
- Joined: Fri Apr 01, 2011 2:14 pm
Re: Adding iterations
Hi Lea,
I don't think there is an easy automated wayo do what you want using runmlwin.
However, given that you say that the MCMC chain has not converged to its posterior distribution after 100,000 iterations, you should discard these iterations anyway since your final inferences should be based only on iterations from the stationary (i.e., converged) joint posterior distribution. One thing you might want to therefore try is to take the final set of parameter values from your existing MCMC chains (i.e., at the 100,000 iteration) and specify them as starting values in a new model where you run for another 100,000 iterations.
Best wishes
George
I don't think there is an easy automated wayo do what you want using runmlwin.
However, given that you say that the MCMC chain has not converged to its posterior distribution after 100,000 iterations, you should discard these iterations anyway since your final inferences should be based only on iterations from the stationary (i.e., converged) joint posterior distribution. One thing you might want to therefore try is to take the final set of parameter values from your existing MCMC chains (i.e., at the 100,000 iteration) and specify them as starting values in a new model where you run for another 100,000 iterations.
Best wishes
George