Independent linear statistics on the cylinders

TL;DR

This paper extends the Skitovich--Darmois theorem to certain group structures, showing that independence of specific linear statistics of independent variables implies Gaussian distributions.

Contribution

It proves an analogue of the Skitovich--Darmois theorem for groups like cylinders and solenoids, identifying conditions under which distributions are Gaussian.

Findings

Linear statistics' independence implies Gaussian distributions.

The result applies to groups like $ extbf{R} imes extbf{T}$ and solenoids.

The characteristic functions of the variables do not vanish.

Abstract

Let either $X=\mathbf{R}\times\mathbf{T}$ or $X=\Sigma_\text{\boldmath $a$}\times\mathbf{T}$, where $\mathbf{R}$ is the additive group of real number, $\mathbf{T}$ is the cycle group and $\Sigma_\text{\boldmath $a$}$ is an $\text{\boldmath $a$}$-adic solenoid . Let $\alpha_{ij}$, where $i, j=1,2,3,$ be topological automorphisms of the group $X$. We prove the following analogue of the well-known Skitovich--Darmois theorem for the group $X$. Let $\xi_j$, where $j=1, 2, 3$, be independent random variables with values in the group $X$ and distributions $\mu_j$ such that their characteristic functions do not vanish. If the linear statistics $L_1=\alpha_{11}\xi_1+\alpha_{12}\xi_2+\alpha_{13}\xi_3$, $L_2=\alpha_{21}\xi_1+\alpha_{22}\xi_2+\alpha_{23}\xi_3$, and $L_3=\alpha_{31}\xi_1+\alpha_{32}\xi_2+\alpha_{33}\xi_3$ are independent, then all $\mu_j$ are Gaussian distributions.