Bayesian model selection for spatial clustering in 3D surveys

Abstract

In this preliminary work on galaxy clustering, we study clusters of arbitrary shape using a 3D galaxy catalog (celestial coordinates and photometric redshifts) derived from Sloan Digital Sky Survey Early Release Data. Spatial influence is modeled using a Markov random field. Comparative model assessment is carried out using an approximation to Bayes factors, viz. the posterior odds of a hypothesis of a given number of clusters versus another. We conclude with a discussion of promising future research directions.