Wednesday 11 May 2011 at 1:10 pm
Bayesian Model Averaging (BMA) allows for any kind of model prior distributions. While the R package BMS has built-in support for several types of commonly used priors, there may be the need for constructing a custom model prior in a particular exercise.
This tutorial is for users which already dispose of some experience in BMS. Beginners might find other tutorials vastly more helpful.
Saturday 06 November 2010 at 12:17 am
This text 'pedagocially' explains how the bms function works code-wise and is intended for people who prefer to program customized adjustments of the bms package.
bms is the workhorse function to do the sampling part of Bayesian Model Averaging in the BMS package. The bms function code in the package includes many different options, usability checks, and is tweaked for computational speed. Although it can be difficult to understand, it basically relies on object-oriented subfunctions that can be easily re-shuffled. In order to demonstrate this, the following paragraphs introduce code that replicate a very basic BMA sampling operation but stripped of all options and speed features.
