Statistics in astronomy

TL;DR

Astronomy heavily relies on statistical methods due to the inaccessibility of direct experimentation, and recent advances address complex data analysis challenges in large-scale astronomical datasets.

Contribution

The paper reviews the historical development and current challenges of astrostatistics, highlighting new statistical methods for analyzing complex astronomical data.

Findings

Resurgence of astrostatistical methods with survey data

Development of models for galaxy clustering and cosmic microwave background

Addressing heteroscedastic errors and censored data in astronomy

Abstract

Perhaps more than other physical sciences, astronomy is frequently statistical in nature. The objects under study are inaccessible to direct manipulation in the laboratory, so the astronomer is restricted to observing a few external characteristics and inferring underlying properties and physics. Astronomy played a profound role in the historical development of statistics from the ancient Greeks through the 19th century. But the fields drifted apart in the 20th century as astronomy turned towards astrophysics and statistics towards human affairs. Today we see a resurgence in astrostatistical activity with the proliferation of survey mega-datasets and the need to link complicated data to nonlinear astrophysical models. Several contemporary astrostatistical challenges are outlined: heteroscedastic measurement errors, censoring and truncation in multivariate databases; time series analysis of variable objects including dynamical models of extrasolar planetary systems; treatments of faint sources and other Poisson processes; the anisotropic spatial point process of galaxy clustering; and model fitting and selection for the cosmic microwave background.