Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given that $X(t)$ is a homogeneous poisson process with arrival rate $\lambda$, how do I perform the transformation:

$$ Y(t) =\int ^t _{t-T} X(\zeta) d\zeta$$

to determine say $P(Y(t) < n)$?

share|cite|improve this question
Got something from the answer below? – Did Mar 28 '13 at 22:26

Separating the contributions of the events of the Poisson process before $t-T$ and between $t-T$ and $t$, one gets $Y(t)=TX(t-T)+Z_t(T)$ where $Z_t(T)$ is independent of $X(t-T)$ and distributed as $Z(T)=\sum\limits_{n\geqslant0}T_n\mathbf 1_{T_n\leqslant T}$, where $(T_n)_{n\geqslant1}$ is the sequence of jump times of a Poisson process with intensity $\lambda$. Equivalently, let $(U_n)_{n\geqslant1}$ denote some i.i.d. random variables uniform on $(0,1)$ and $N_T$ an independent Poisson random variable with mean $\lambda T$, then $$ Z(T)=T\sum_{k=1}^{N_T}U_k. $$ Summing up, considering a second Poisson random variable $\bar N_{t-T}$ with mean $\lambda(t-T)$ and independent of the rest, one gets $$ Y(t)=T\bar N_{t-T}+T\sum_{k=1}^{N_T}U_k. $$ Turning to distributions, the Laplace transform reads, for every nonnegative $s$, $$ \mathbb E(\mathrm e^{-sY(t)})=\mathbb E(\mathrm e^{-sTN_{t-T}})\cdot\mathbb E(u_T(s)^{N_T}), $$ where, for every $s$ and $t$, $$ u_T(s)=\mathbb E(\mathrm e^{-sTU})=\frac{1-\mathrm e^{-sT}}{sT},\qquad \mathbb E(\mathrm e^{-sTN_t})=\mathrm e^{-\lambda t(1-\mathrm e^{-sT})}. $$ Thus, $$ \mathbb E(\mathrm e^{-sY(t)})=\exp\left(-\lambda (t-T)(1-\mathrm e^{-sT})-\lambda T\left(1-\frac{1-\mathrm e^{-sT}}{sT}\right)\right), $$ which can be somewhat simplified, but not much. Moments, on the other hand, are easier, for example, $$ \mathbb E(Y(t))=T\lambda(t-T)+T\lambda T\tfrac12=\lambda T(t-\tfrac12T). $$

share|cite|improve this answer
How are your two dexcriptions for the process $Z(T)$ equivalent? The first is very intuitive, attained by constructing the "area" under a Poisson process by adding rectangles, but I can't see how the $Uni((0,1))$ distribution arises in this context. – madison54 Mar 28 '13 at 17:16
In a homogenous Poisson process, conditionally on $N_T=n$, the set of jump times before time $T$ is distributed like the set of values of $n$ i.i.d. random variables uniformly distributed on $(0,T)$. And any uniform random variable on $(0,T)$ coincides with $T$ times a uniform random variable on $(0,1)$. – Did Mar 28 '13 at 22:25
I thought the jump times are distributed as $T_n = t_1 + \dots + t_n$ with $t_i$ i.i.d. Exp($\lambda$) and thus follows a Gamma distribution. Where is my mistake? Does it have to do with conditioning on a fixed number of arrival times in a fixed interval? – madison54 Mar 29 '13 at 9:04
Yes, the "uniform" description is valid conditionally on the number of jumps in (0,T). – Did Mar 29 '13 at 9:45

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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