0
$\begingroup$

The problem below is Exercise 3.8 of Last and Penrose Lectures on Poisson Processes. I have been thinking for the better part of the day about it but couldn't write down even one insightful line. I am especially puzzled about the connection with the Superposition theorem.

Problem: Let $\lambda$ be an $s$-finite measure on $( \mathbb{X}, \mathcal{X})$ and let $\eta$ be a Poisson point process with intensity measure $\lambda$. Let $f:\boldsymbol{N} \to \mathbb{R}_+ $ measurable, be such that $\mathbb{E}(f(\eta)) < \infty$ and further let $\eta'$ be a Poisson point process with intensity measure $\lambda'$ such that $\lambda= \lambda' + \nu$ where $\nu$ is a finite measure.

Apply the Superposition Theorem in order to show that $\mathbb{E}(f ( \eta')) < \infty$


Note: $\boldsymbol{N}$ denotes the set of point measures

My approach: Since the Superposition theorem is mentioned in the exercise my only idea was to introduce another Poisson point process $X$ with intensity $\nu$ that is independent of $\eta'$, by the Superposition theorem I could then conclude that $X + \eta'$ is a Poisson point process with intensity $\lambda' + \nu = \lambda$.

Update: I have managed to write down some additional lines, namely
Let $\eta \sim $PPP($\lambda$) and $\eta' \sim$ PPP($\lambda'$), choose $X \sim$PPP($\nu$) such that $X$ is independent of $\eta'$, by the Superposition theorem for PPP we obtain that $$ \eta' + X \sim \text{PPP}(\lambda' + \nu)= \text{PPP}(\lambda) \sim \eta $$ In particular for $f: \boldsymbol{N} \to \mathbb{R}_+$ measurable we have that $$ \mathbb{E}(f( \eta' + X)) = \mathbb{E}(f ( \eta)) < \infty $$ Certainly we have $\eta' \leq \eta' + X$, I need to rigorously justify why $$ \mathbb{E}(f ( \eta')) \leq \mathbb{E}(f( \eta' +X)) \tag{*} $$ The reason for this must be connected to the fact that we have $\lambda = \lambda' + \nu$ where both $\lambda, \lambda'$ are $s$-finite and $\nu$ is finite. Since $\nu$ is a finite measure I can say that $\lambda' \leq \lambda' + \nu$ and on (*) the integral on the RHS measures all elements of $\boldsymbol{N}$ that get measured on the LHS but finitely many more because $\nu$ is finite, thus * holds.

Questions: Is my reasoning correct? Is there a way how to rigorously justify * ?

$\endgroup$
0
$\begingroup$

As declared in the OP we let $ \eta \sim $PPP($\lambda$) and $\eta' \sim$ PPP( $\lambda'$) where $\lambda, \lambda'$ are $\sigma$-finite measures. We then choose $X \sim $PPP($\nu$) where $\nu$ a finite measure such that $X$ is independent of $\eta'$.

By the Superposition theorem it follows that $\eta' + X \sim$ PPP( $\lambda' + \nu$)=PPP($\lambda) \sim \eta$ and in particular we have for $f: \boldsymbol{N} \to \mathbb{R}_+$ measurable that $\mathbb{E}(f( \eta' + X))= \mathbb{E}(f( \eta))< \infty$. Furthermore we have $$ \infty > \mathbb{E}(f( \eta' + X)) \geq \mathbb{E}(f( \eta') 1_{X=0})\overset{1}=\mathbb{E}(f( \eta')) \mathbb{P}(X=0) \overset{2}= \mathbb{E}(f( \eta')) e^{- \lambda} \\ \overset{3}\implies \mathbb{E}(f( \eta'))< \infty$$

1) $X$ is independent of $\eta'$
2) $X \sim $PPP($\lambda)$
3) $\lambda$ is a finite measure and thus $e^{-\lambda}>0$

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.