New Statistical Methods for Analysis of Large Surveys Distributions and Correlations.pdfVIP

New Statistical Methods for Analysis of Large Surveys: Distributions and Correlations Vahe’ Petrosian Center for Space Science and Astrophysics, Varian 302c, Stanford University, Stanford, CA 94305-4060 1 0 0 ABSTRACT 2 c e The aim of this paper is to describe new statistical methods for determination of the D correlations among and distributions of physical parameters from a multivariate data 9 1 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 1 analysis of combined data from many surveys with different and varied observational v 7 selection criteria. For clarity I will use the luminosity function of AGNs and its evolution 6 to demonstrate the methods. I will first describe the general features of data truncation 4 2 and present a brief review of past methods of analysis. Then I will describe the new 1 methods and results from simulations testing their accuracy. Finally I will present the 1 0 results from application of the methods to a sample of quasars. / h p - o 1. INTRODUCTION r t s a One of the important ways of testing the models of AGN, or any other astrophysical source, : v is through the investigations of the distributions, ranges, and more importantly the correlations i X among, the relevant physical characteristics, such as luminosity, spectrum,