Exponential densities and compound Poisson measures

TL;DR

This paper develops a new method to analyze the asymptotic behavior of convolutions and compound Poisson measures for a class of radial decreasing densities on , especially those not convolution equivalent, with applications to exponential-product densities.

Contribution

It introduces a novel approach to estimate convolutions and compound Poisson densities beyond the convolution equivalent case, focusing on exponential-product densities with polynomial terms.

Findings

Established asymptotic estimates for convolutions of non-convolution equivalent densities.

Derived density estimates involving generalized Bessel functions for polynomial-modified exponential densities.

First analysis of compound Poisson measures for these complex density classes.

Abstract

We prove estimates at infinity of convolutions $f^{n\star}$ and densities of the corresponding compound Poisson measures for a class of radial decreasing densities on $\mathbb{R}^d$, $d \geq 1$, which are not convolution equivalent. Existing methods and tools are limited to the situation in which the convolution $f^{2\star}(x)$ is comparable to initial density $f(x)$ at infinity. We propose a new approach which allows one to break this barrier. We focus on densities which are products of exponential functions and smaller order terms -- they are common in applications. In the case when the smaller order term is polynomial estimates are given in terms of the generalized Bessel function. Our results can be seen as the first attempt to understand the intricate asymptotic properties of the compound Poisson and more general infinitely divisible measures constructed for such densities.