In Bayesian statistics, models are evaluated on the basis of data by comparing their marginal likelihoods conditional on the data. This package helps compute models' marginal likelihoods using samples from the models' posterior parameter distributes (the inputs are similar to those of the harmonic mean estimator). The computation itself is done via "arrogance sampling", a kind of nonparametric (histogram-based) importance sampling.
| Version: | 0.2 |
| Depends: | R (≥ 2.7.0) |
| Published: | 2011-01-04 |
| Author: | Benedict Escoto |
| Maintainer: | Benedict Escoto <favir at emerose.org> |
| License: | GPL (≥ 2) |
| CRAN checks: | margLikArrogance results |
| Package source: | margLikArrogance_0.2.tar.gz |
| MacOS X binary: | margLikArrogance_0.2.tgz |
| Windows binary: | margLikArrogance_0.2.zip |
| Reference manual: | margLikArrogance.pdf |
| Vignettes: |
Technical paper explaining and justifying method Usage example in a simple Bayesian model choice problem |
| Old sources: | margLikArrogance archive |