Importance sampling
Posted: Wed Sep 23, 2009 2:43 am
Hi Marc,
I already asked a similar question before, but maybe in the wrong context. I couldn't find any documentation on the importance sampling neither on the web nor in your thesis. The way I used the importance sampling was by adding a new run, setting RunType to 'Importance sampling' and then start it as a joint run with the previous 'Optimization' run. Then I assumed that the models created from this run sample the approximate PPD of the previous run, as described in Sambridge1999b. Thus I should be able to calculate bayesian Integrals from the ensemble parameters of the 'Importance sampling' run. Are these assumptions right?
Cheers,
Yannik
I already asked a similar question before, but maybe in the wrong context. I couldn't find any documentation on the importance sampling neither on the web nor in your thesis. The way I used the importance sampling was by adding a new run, setting RunType to 'Importance sampling' and then start it as a joint run with the previous 'Optimization' run. Then I assumed that the models created from this run sample the approximate PPD of the previous run, as described in Sambridge1999b. Thus I should be able to calculate bayesian Integrals from the ensemble parameters of the 'Importance sampling' run. Are these assumptions right?
Cheers,
Yannik