I am self-studying probability theory and currently the Poisson point process (PPP) gives me hell, firstly because the definition of a point process in general and PPP in particular seems rather cumbersome at first to tenth glance and secondly because there are no examples in the script I use and hardly any on the net. So I constructed my own example, regarding which I have some questions... Sorry for the long intro, here comes the math:

Suppose we distribute $n$ points randomly and uniformly on the interval $[0, n]$. Thus we have $X_1, ..., X_n \sim Unif([0, n])$. For each $\omega \in \Omega$ we have $N(\omega) = \sum_{i = 1}^{10} \delta_{X_i(\omega)}$ as a random counting measure on $[0, n]$. Obviously

$\mathbb{P}[N(\omega)[B] = k] = \binom{n}{k}(\frac{\lambda[B]}{n})^k(\frac{\lambda[B^c]}{n})^{n-k}\sim Bin(n, \frac{\lambda[B]}{n})$.

Now since $Bin(n, \frac{\lambda[B]}{n}) \rightarrow_{n\to\infty} Pois(\lambda[B])$, the limit process on $\mathbb{R}_+$ with infinitely many points, should be a PPP. Is this correct so far? If so, how to prove independence on disjoint subintervals?

Now the problem is, that there is no uniform distribution on $\mathbb{R}_+$, so the process with underlying random variables can only be defined as a limit process. I wondered if this may be the reason, why a point process is defined as a measurable map from $\Omega$ to the set of locally finite counting measures. Is this the case?

I would also greatly appreciate other examples of PPPs "built from scratch", ie with explicit definition of the map $N:\Omega\rightarrow\mathfrak{N}$, where $\mathfrak{N}$ is the set of locally finite counting measures. Thanks for help!

Edit: Maybe I found another way to do it, please correct, if there are any errors.

Let $A_n := [n, n+1)$, thus $\mathbb{R}_+ = \sqcup_{n \in \mathbb{N}_0} A_n$. Let $(X_{n, j})_{n, j \in \mathbb{N}_0} \sim Unif([0, 1))$ be i.i.d., $(Y_n)_{n \in \mathbb{N}_0} \sim Pois(1)$ be i.i.d.. Define

$N_n: \Omega \rightarrow \mathfrak{N}(A_n, \mathfrak{B}(A_n)), \omega \mapsto \sum_{j=1}^{Y_n(\omega)}\delta_{(X_{n, j})+n)}$, where again $\mathfrak{N}(A_n, \mathfrak{B}(A_n))$ is the set of locally finite counting measures on $(A_n, \mathfrak{B}(A_n))$. Then

$N:=\sum_{n\in\mathbb{N}_0}N_n = \sum_{n\in\mathbb{N}_0} \delta_{X_n}$ a.s.

does the job.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.