An unexpected encounter with Cauchy and Lévy

DOI10.1214/15-AOS1407zbMATH Open1349.62036arXiv1505.01957OpenAlexW2964112220WikidataQ56115768 ScholiaQ56115768MaRDI QIDQ342671FDOQ342671

Authors: Natesh S. Pillai, Xiao-Li Meng 

Publication date: 18 November 2016

Published in: The Annals of Statistics (Search for Journal in Brave)

Abstract: The Cauchy distribution is usually presented as a mathematical curiosity, an exception to the Law of Large Numbers, or even as an "Evil" distribution in some introductory courses. It therefore surprised us when Drton and Xiao (2016) proved the following result for $m=2$ and conjectured it for $mge3$. Let $X=(X_1,\mathrm{...},X_m)$ and $Y=(Y_1,\mathrm{...},Y_m)$ be i.i.d $N(0,Sigma)$, where $Sigma=sigm{a}_{ij}ge0$ is an $mimesm$ and extit{arbitrary} covariance matrix with $sigm{a}_{jj}>0$ for all $1leqjleqm$. Then ${\displaystyle }$Z = sum_{j=1}^m w_j frac{X_j}{Y_j} sim mathrm{Cauchy}(0,1),${\displaystyle }$ as long as $w=(w_1,\mathrm{...},w_m)$ is independent of $(X,Y)$, $w_{j}ge0,j=1,\mathrm{...},m$, and $su{m}_{j=1}^m{w}_j=1$. In this note, we present an elementary proof of this conjecture for any $mgeq2$ by linking $Z$ to a geometric characterization of Cauchy(0,1) given in Willams (1969). This general result is essential to the large sample behavior of Wald tests in many applications such as factor models and contingency tables. It also leads to other unexpected results such as ${\displaystyle }$ sum_{i=1}^msum_{j=1}^m frac{w_iw_jsigma_{ij}}{X_iX_j} sim { ext{L'{e}vy}}(0, 1). ${\displaystyle }$ This generalizes the "super Cauchy phenomenon" that the average of $m$ i.i.d. standard L'evy variables (i.e., inverse chi-squared variables with one degree of freedom) has the same distribution as that of a single standard L'evy variable multiplied by $m$ (which is obtained by taking $w_j=1/m$ and $Sigma$ to be the identity matrix).

Full work available at URL: https://arxiv.org/abs/1505.01957

Recommendations

• Cauchy, normal and correlations versus heavy tails
• Sample means of independent standard Cauchy random variables are standard Cauchy: a new approach
• scientific article; zbMATH DE number 3986377
• scientific article; zbMATH DE number 3907464
• On characterization of the Cauchy law by sample mean properties

zbMATH Keywords

Wald testCauchy distributionratio of GaussiansLévy distribution

Mathematics Subject Classification ID

Exact distribution theory in statistics (62E15) Multivariate distribution of statistics (62H10) Hypothesis testing in multivariate analysis (62H15)

Cited In (14)

• Centers of probability measures without the mean
• Cauchy Combination Test: A Powerful Test With Analytic p-Value Calculation Under Arbitrary Dependency Structures
• Title not available (Why is that?)
• Pseudorandom processes
• A robust and fast two-sample test of equal correlations with an application to differential co-expression
• Cauchy, normal and correlations versus heavy tails
• Sample means of independent standard Cauchy random variables are standard Cauchy: a new approach
• Title not available (Why is that?)
• Large-Scale Hypothesis Testing for Causal Mediation Effects with Applications in Genome-wide Epigenetic Studies
• Beyond the two-trials rule: Type-I error control and sample size planning with the sceptical $p$-value
• Global-local mixtures: a unifying framework
• Stable Combination Tests
• Heavy-tailed distributions, correlations, kurtosis and Taylor’s Law of fluctuation scaling
• Title not available (Why is that?)

This page was built for publication: An unexpected encounter with Cauchy and Lévy

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q342671)