Nonhomogeneous Poisson process

107 Views Asked by At

Let $\lambda:[0,\infty)\rightarrow [0,\infty)$ be a continuous function and $N$ be a Poisson process with rate $1$. Define $\Lambda(t)=\int_0^t{\lambda(x)dx}$ then how do we prove that $$ lim_{s\rightarrow{0}}{\frac{1-P(N(\Lambda(t+s))-N(\Lambda(t))=0)}{t}}=\lambda(t)? $$

I found the following result in many textbooks:

$N(\Lambda(t+s))-N(\Lambda(t))$ is a Poisson random variable with parameter $\int_t^{t+s}\lambda(x)dx$ and if we use this result then to prove the above limit is almost obvious. But I am not sure how to prove the second result or is there any way to get the above limit without using the second result?

1

There are 1 best solutions below

3
On BEST ANSWER

Yes the result is obvious since by definition $N(u+v)-N(u)$ is Poisson with parameter $v$, for every nonnegative $u$ and positive $v$, hence $$ \lim_{v\to0}\frac{P(N(u+v)-N(u)\geqslant1)}v=\lim_{v\to0}\frac{1-\mathrm e^{-v}}v=1,$$ which, for $u=\Lambda(t)$ and $v=\Lambda(t+s)-\Lambda(t)$, implies $$ \lim_{s\to0}\frac{P(N(\Lambda(t+s))-N(\Lambda(t))\geqslant1)}{\Lambda(t+s)-\Lambda(t)}=1, $$ and the result in the question then follows from the other limit $$\lim_{s\to0}\frac{\Lambda(t+s)-\Lambda(t)}s=\lambda(t). $$