Overview

HTTPS SSH

dilutBMS2

dilutBMS2 provides Bayesian Model Averaging for models with multicollinear regressors by including alternative priors such as different sets of Heredity priors introduced by Chipman (1996) and the Tessellation prior presented by George (2010). It is a fork of the BMS package for R written by Stefan Zeugner and Martin Feldkircher.

Installation

Install using the devtools package:

library(devtools)
install_bitbucket("dilutbms2", "matmo")

Usage

For the Tessellation prior set

bms(..., mcmc="tess", pen=2)

the Strong Heredity prior is implemented in the default BMS package and can be accessed via

bms(..., mcmc="bd.int")

alternatively a Weak Heredity Prior can be specified, with a penalty term. The penalty puts a logarithmic penalty depending on the number of sampled Interaction terms and missing parent variables

bms(..., mprior="uniform.wh", mcmc="bd", pen=0.2)

the general Dilution Prior can be accessed through

bms(..., mprior="dilut")