Customizing Priors for BMS

The BMS package (as from version 0.3.0, ) allows to user-define coefficent and model priors as well as MCMC samplers.

Custom model priors

• For an introduction to user-defining custom model priors, consider this tutorial.
• In addition, the manual entry on the class mprior could be helpful.
• The built-in model prior functions are described in appendix of the PDF tutorial Bayesian Model Averaging with BMS.

Custom coefficient priors

• The built-in coefficent priors are described in appendix of the PDF tutorial Bayesian Model Averaging with BMS, and the manual entry on the class gprior has some additional information.
• Usage examples will appear on the BMS Blog by late May 2011.

User-defined MCMC samplers for model sampling

• bms currently offers two built-in Metropolis-Hastings samplers for model sampling (cf. argument mcmc in help(bms)).
• Usage examples will appear on the BMS blog. In the meantime the function .fls.samp in the BMS package may serve as an inspiration.