English

A Cram\'er-Wold theorem for elliptical distributions

Probability 2023-03-10 v2 Statistics Theory Statistics Theory

Abstract

According to a well-known theorem of Cram\'er and Wold, if PP and QQ are two Borel probability measures on Rd\mathbb{R}^d whose projections PL,QLP_L,Q_L onto each line LL in Rd\mathbb{R}^d satisfy PL=QLP_L=Q_L, then P=QP=Q. Our main result is that, if PP and QQ are both elliptical distributions, then, to show that P=QP=Q, it suffices merely to check that PL=QLP_L=Q_L for a certain set of (d2+d)/2(d^2+d)/2 lines LL. Moreover (d2+d)/2(d^2+d)/2 is optimal. The class of elliptical distributions contains the Gaussian distributions as well as many other multivariate distributions of interest. Our theorem contrasts with other variants of the Cram\'er-Wold theorem, in that no assumption is made about the finiteness of moments of PP and QQ. We use our results to derive a statistical test for equality of elliptical distributions, and carry out a small simulation study of the test, comparing it with other tests from the literature. We also give an application to learning (binary classification), again illustrated with a small simulation

Keywords

Cite

@article{arxiv.2206.13612,
  title  = {A Cram\'er-Wold theorem for elliptical distributions},
  author = {Ricardo Fraiman and Leonardo Moreno and Thomas Ransford},
  journal= {arXiv preprint arXiv:2206.13612},
  year   = {2023}
}

Comments

20 pages; 7 figures

R2 v1 2026-06-24T12:06:00.804Z