salleuska / CPLogit

Supplementary code for the paper "Centered Partition Processes: Informative Priors for Clustering"

Supplementary code for Bayesian linear and logistic regression using Centered Partition processes

Code to reproduce simulation results presented in
Centered Partition Processes: Informative Priors for Clustering (with Discussion)
by Sally Paganin, Amy H. Herring, Andrew F. Olshan, David B. Dunson
Bayesian Analysis. 16(1), 301-370, (March 2021) [ Link | arXiV preprint ]

Installation

Be sure to have installed the following R packages

	install.packages("Rcpp")
	install.packages("RcppArmadillo")
	install.packages("devtools")
	install.packages("ggplot2")
	install.packages("reshape2")
	install.packages("magrittr")
	install.packages("dplyr")
	install.packages("sdols")
	install.packages("mcclust")
	install.packages("mvtnorm")

Download and install the package CPLogit using devtools

library(devtools)
devtools::install_github("salleuska/CPLogit", subdir="CPLogit")

or if you download the repo

## How to install 
library(devtools)
library(Rcpp)
clean_dll()

## to export Rcpp functions
compileAttributes()
build()
install()
document()

Reproduce plots in Section 3.3

To reproduce code of the prior probability distribution over the set partition space induced by the CP process, use script Sec3.3_Prior_graphs.R.

Reproduce simulation results

Simulate data used in the Sec 5.2 and in Sec 1 of the Supplementary Materials

Rscript simulation/0_generateData_linearSimulation.R
Rscript simulation/0_generateData_logisticSimulation.R

Run models and reproduce plots in Sec 5.2 and in Sec 1 of the Supplementary Materials

Rscript simulation/1_linearSimulation.R
Rscript simulation/1_logisticSimulation.R