ex2.draw: MCMC samples of Bayesian cluster model for a simulated...

In muschellij2/mcclust.ext: Point estimation and credible balls for Bayesian cluster analysis

Description

MCMC samples of clusterings from a Dirichlet process scale-location mixture model with normal components fitted to a simulated dataset, see ex2.data. True clusters are located at (+/- 2, +/- 2) with a standard deviation of 1, 0.5, 1, and 1.5 in the first, second, third, and fourth quadrant respectively.

Usage

data(ex2.draw)

Format

The matrix ex2.draw has 10,000 rows and 200 columns, with each row representing a MCMC posterior sample of the clustering of the 200 data points contained in ex2.data.

Source

Wade, S. and Ghahramani, Z. (2015) Bayesian cluster analysis: Point estimation and credible balls. Submitted. arXiv:1505.03339.

Examples

data(ex2.data)
data(ex2.draw)
x=data.frame(ex2.data[,c(1,2)])
cls.true=ex2.data$cls.true
plot(x[,1],x[,2],xlab="x1",ylab="x2")
k=max(cls.true)
for(l in 2:k){
points(x[cls.true==l,1],x[cls.true==l,2],col=l)}

# Find representative partition of posterior
psm=comp.psm(ex2.draw)
ex2.VI=minVI(psm,ex2.draw,method=("all"),include.greedy=TRUE)
summary(ex2.VI)
plot(ex2.VI,data=x)

# Uncertainty in partition estimate
ex2.cb=credibleball(ex2.VI$cl[1,],ex2.draw)
summary(ex2.cb)
plot(ex2.cb,data=x)

muschellij2/mcclust.ext documentation built on May 26, 2019, 9:36 a.m.