Bayesclust: An R package for testing and searching for significant clusters

Abstract

The detection and determination of clusters has been of special interest among researchers from different fields for a long time. In particular, assessing whether the clusters are significant is a question that has been asked by a number of experimenters. In Fuentes and Casella (2009), the authors put forth a new methodology for analyzing clusters. It tests the hypothesis H0: κ = 1 versus H1: κ = k in a Bayesian setting, where κ denotes the number of clusters in a population. The bayesclust package implements this approach in R. Here we give an overview of the algorithm and a detailed description of the functions available in the package. The routines in bayesclust allow the user to test for the existence of clusters, and then pick out optimal partitionings of the data. We demonstrate the testing procedure with simulated datasets.