Statistical Methods for Astronomy

TL;DR

This review covers statistical concepts, methods, and tools essential for analyzing modern astronomical data, including inference techniques, resampling, model selection, and specialized statistical procedures tailored for astronomy.

Contribution

It provides a comprehensive overview of statistical methods and resources, including R software and packages, specifically adapted for astronomical data analysis.

Findings

Resampling methods like bootstrap are valuable when distribution functions are unknown.

Various nonparametric and multivariate methods are applicable to astronomical datasets.

R software and extensive packages facilitate advanced statistical analysis in astronomy.

Abstract

This review outlines concepts of mathematical statistics, elements of probability theory, hypothesis tests and point estimation for use in the analysis of modern astronomical data. Least squares, maximum likelihood, and Bayesian approaches to statistical inference are treated. Resampling methods, particularly the bootstrap, provide valuable procedures when distributions functions of statistics are not known. Several approaches to model selection and good- ness of fit are considered. Applied statistics relevant to astronomical research are briefly discussed: nonparametric methods for use when little is known about the behavior of the astronomical populations or processes; data smoothing with kernel density estimation and nonparametric regression; unsupervised clustering and supervised classification procedures for multivariate problems; survival analysis for astronomical datasets with nondetections; time- and frequency-domain times series analysis for light curves; and spatial statistics to interpret the spatial distributions of points in low dimensions. Two types of resources are presented: about 40 recommended texts and monographs in various fields of statistics, and the public domain R software system for statistical analysis. Together with its \sim 3500 (and growing) add-on CRAN packages, R implements a vast range of statistical procedures in a coherent high-level language with advanced graphics.