Department of Statistics Unitmark
Dietrich College of Humanities and Social Sciences

Nonparametric Density Estimation and Clustering in Astronomical Sky Surveys

Publication Date

April, 2004

Publication Type

Tech Report

Author(s)

Woncheol Jang

Abstract

We present a nonparametric method for galaxy clustering in astronomical sky surveys. We show that the cosmological definition of clusters of galaxies is equivalent to density contour clusters (Hartigan, 1975) \(S_c = \{ f > c \}\) where \(f\) is a probability density function. The plug-in estimator \(\hat S_c =\{ \hat f > c \}\) is used to estimate \(S_c\) where \(\hat f\) is the multivariate kernel density estimator. To choose the optimal smoothing parameter, we use cross-validation and the plug-in method and show that cross-validation method outperforms the plug-in method in our case. A new cluster catalogue based on the plug-in estimator is compared to existing cluster catalogs, the Abell and EDCCI. Our result is more consistent with the EDCCI than with the Abell, which is the most widely used catalogue. We present a practical algorithm for local smoothing and use the smoothed bootstrap to asses the validity of clustering results.

(Revised 10/04)