Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Investigate and implement bridge sampling for marginal likelihood computation #884

Open
mhtess opened this issue Sep 21, 2017 · 0 comments

Comments

@mhtess
Copy link
Contributor

mhtess commented Sep 21, 2017

It's recently been suggested that bridge sampling is a reliable and relatively straightforward algorithm for obtaining the marginal likelihood for purposes of model comparison. It may be sufficiently general to cover many of our use cases for model comparison of psychologically interesting models

There is an R package, but you need to specify a log_posterior function that can map parameter values to unnormalized log posterior values (hence, also a likelihood of the data). Thus, for this, you must have an analytic likelihood function, which is not the situation we are usually in with recursive models like RSA. They base their approach on Meng & Wong (1996)

In theory the R package could be harnessed using RWebPPL, but in rwebppl's current implementation, the model must be recompiled every time you have new input data (i.e., a new proposal to the parameters)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant