New Statistical Methods for Analysis of Large Surveys: Distributions and Correlations

Abstract

The aim of this paper is to describe new statistical methods for determination of the correlations among and distributions of physical parameters from a multivariate data with general and arbitrary truncations and selection biases. These methods, developed in collaboration with B. Efron of Department of Statistics at Stanford, can be used for analysis of combined data from many surveys with different and varied observational selection criteria. For clarity we will use the luminosity function of AGNs and its evolution to demonstrate the methods. We will first describe the general features of data truncation and present a brief review of past methods of analysis. Then we will describe the new methods and results from simulations testing their accuracy. Finally we will present the results from application of the methods to a sample of quasars.