Poisson counting measure

668 Views Asked by At

I'm working through the proof of Theorem VI.2.17 in the book "Probability and Stochastics" by Cinlar, and am having some problems understanding it (or maybe there are some conditions missing?)

Let $(E,\mathcal{E})$ be a measurable space, and let N be a Poisson random measure on it. Suppose that its mean (measure) $\nu$ is $\Sigma$-finite. Then, N is a random counting measure if and only if $\nu is diffuse.

I'll recall some of the definitions here, in case the reader is not familiar with it (since it might not be standard definitions).

A measure $\nu$ is diffuse if $\nu(\{x\}) =0$ for every $x \in E$.

A measure $\nu$ is $\Sigma$-finite if there exists a sequence of finite measures $\nu_n$ such that $\nu = \sum_n \nu_n$.

A random measure N is called a random counting measure if, for almost every $\omega$, the measure $A \mapsto M(\omega,A)$ is purely atomic and its every atom has weight one.

The necessity part of the proof is quite simple, we just show that $\nu(\{x\}) = 0$ for fixed (arbitrary) $x \in E$, which follows from some simple calculation for the random counting measure N.

However, in the sufficiency part I'm not sure if every step is correct. The proof goes as follows:

We assume that $\nu$ is diffuse (and $\Sigma$-finite). Because the probability law of a Poisson random measure is determined by its mean measure, we may assume that N is constructed by $$N = \sum_{i=1}^\infty N_n, \qquad N_n = \sum_{i=1}^{K_n(\omega)} 1_A (X_{n,i}),$$ where the $X_i$ are i.i.d and K is a Poisson random variable with mean $\nu$. (I'm not actually sure that this holds without any further assumptions, I have a feeling we may need to restrict it to countably generated $\mathcal{E}$). Note that the collection $X = \{X_{n,i}: n \geq 1, i \geq 1\}$ is independent, and every member has a diffuse distribution.

Take some pair of distinct indices $(n,i)$ and $(m,j)$. Then $P = (X_{ni} = X_{mj}) = 0$ (I'm also not sure this part holds without any further assumptions). Then, the countable union of all these events is also negligible, and hence N is a counting measure.

I'm looking for explanations why these parts of the proof are correct - or that further conditions are needed.