galaxy.draw: MCMC samples of a Bayesian cluster model for the galaxy...

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 the galaxies dataset.

Usage

data(galaxy.draw)

Format

The matrix galaxy.draw has 10,000 rows and 82 columns, with each row representing a MCMC posterior sample of the clustering of the 82 data points.

Source

Roeder, K. (1990) Density estimation with confidence sets exemplified by superclusters and voids in the galaxies, Journal of the American Statistical Association, 85: 617-624.

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

Examples

data(galaxy.fit)
x=data.frame(x=galaxy.fit$x)
data(galaxy.pred)
data(galaxy.draw)

# Find representative partition of posterior
psm=comp.psm(galaxy.draw)
galaxy.VI=minVI(psm,galaxy.draw,method=("all"),include.greedy=TRUE)
summary(galaxy.VI)
plot(galaxy.VI,data=x,dx=galaxy.fit$fx,xgrid=galaxy.pred$x,dxgrid=galaxy.pred$fx)

# Uncertainty in partition estimate
galaxy.cb=credibleball(galaxy.VI$cl[1,],galaxy.draw)
summary(galaxy.cb)
plot(galaxy.cb,data=x,dx=galaxy.fit$fx,xgrid=galaxy.pred$x,dxgrid=galaxy.pred$fx)

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