My questions will be given at the end, let me just give some definitions first.

The counting process $\{ N(t), t \geq 0 \} $ is said to be a non homogenous Poisson process with intensity function $\lambda(t), t \geq 0$ if:

$N(0)= 0$

$\{ N(t),t \geq 0 \}$ has independent increments

$P\{ N(t+h) - N(t) \geq 2 \} = o(h)$

$P\{ N(t+h) - N(t) =1 \} =\lambda(t)h + o(h)$

Time sampling and ordinary poisson process generates a nonhomgenous poission process. Let $\{N(t),t\geq 0 \} $ be a poisson process with rate $\lambda$ and suppose that an event occurring at time $t$ is , independently of what has occurred prior to $t$, counted with probability $p(t)$,then the counting process $\{N_c(t), t \geq 0 \} $ is a nonhomogenous poisson process with intensity function $\lambda(t) = \lambda\cdot p(t)$

The book now shows that a time sampling satisfies the axioms above.verifying the last axiom we calculate

$P\{ N_c(t+h) - N_c(t) =1 \}=P\{ N_c(t,t+h)=1\} = \cdots =P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \}\lambda h + o(h) = p(t)\lambda h + o(h) $

Now, what i don't understand is how $P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \} = p(t)$, the way the book defined $p(t)$ (see above). First of all we are conditioning on that the regular poisson process is one, and also the above says that there is 1 event happening in the intervall $(t,t+h)$ , but $p(t)$ is the probability that an event occur at time $t$.

Could someone clarify this?


Events are occurring under the homogeneous process $N$ at rate $\lambda$. Each such event, occurring at time $t$ say, is then either counted in the $N_c$ process, with probability $p(t)$, or ignored with probability $1-p(t)$. So it's not correct to say that $p(t)$ is the probability that an event occurs at time $t$ in process $N_c$.

Mathematically, we can say:

$$p(t) = P(\text{Event at time $t$ is counted}\mid\text{Event occurs under process $N$ at time $t$}).$$

This is close to the line in the proof you refer to but there is a time interval of size $h$, so it is only an approximation

$$P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \} = p(t).$$

If the event occured at time $t_0 \in [t,t+h]$ then the true value for the RHS would be $p(t_0)$. But for small $h$, $p(t)$ is a reasonable approximation. Also, the "error" in this approximation is accounted for in the term $o(h)$ allowing the use of the equality sign.

  • $\begingroup$ is this the right reasoning? : $P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \ \} = p(t) + g(h)$ where $g(h)$ is a function such that $lim_{h \rightarrow 0} g(h) = 0$? @Mick A, because as h get smaller and smaller then we finally get $p(t)$ $\endgroup$ – Danny Jun 17 '15 at 9:40
  • $\begingroup$ @Danny Almost. The limit has to be stronger than that though. We need $\lim_{h\to 0} (g(h)/h) = 0$. So $g(h)$ has to approach $0$ much faster than $h$ itself does. For example, from the definition, we have $P\{N(t+h)-N(t)=1\}=\lambda(t)h+o(h)$. Here, as $h\to 0$, both $\lambda(t)h$ and $o(h)$ approach $0$ but $o(h)$ does so much faster than does $\lambda(t)h$. $\endgroup$ – Mick A Jun 17 '15 at 12:57
  • $\begingroup$ Yes i know what you mean, but i think you forgot that $P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \} = p(t)$ above must be multiplied with $\lambda h$ (see above), so what I mean is that it should be $P\{N_c(t,t+h) = 1 \mid N(t,t+h) = 1 \}\lambda h = (p(t) + g(h))\lambda h$. and therefore $t(h) = g(h)\lambda h = o(h)$, then it is enough that $g(h)$ goes to zero. However when i think about it why should the added contribution to $p(t)$ be a function of 'h' so my reasoning seems not be right. @Mick A $\endgroup$ – Danny Jun 17 '15 at 16:09
  • $\begingroup$ @Danny I don't think $P\{N_c(t,t+h)=1\mid N(t,t+h)=1\}=p(t)+g(h)$. It is just $P\{N_c(t,t+h)=1\mid N(t,t+h)=1\}=p(t)$ because you are given that there is a single event ($N(t,t+h)=1$) and we know that with probability $p(t)$ it is counted in process $N_c$. (If I understand you correctly, this is kind of what you're saying in your last sentence.) $\endgroup$ – Mick A Jun 17 '15 at 17:15
  • $\begingroup$ .but $P\{N_c(t,t+h)=1\mid N(t,t+h)=1\}$is not $p(t)$, but as you said before, it is an approximation, so what is actually $P\{N_c(t,t+h)=1\mid N(t,t+h)=1\}$. @Mick A $\endgroup$ – Danny Jun 17 '15 at 17:20

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.