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

I am having trouble in understanding this definition of Poisson process.

Let $S$ be a random discrete subset of points of $\mathbb{R}^d$ and let $\lambda >0$.

  1. A partition $\mathcal{A}$ of $\mathbb{R}^d$ with $A\in \mathcal{A}$ measurable and $l(A)<\infty$.

  2. Independent Poisson random variables $Y_A\sim\text{Poisson}(\lambda l(A))$.

  3. A family $((U_{A,j}, j\ge 1) A\in \mathcal{A})$, where $U_{A,j}\sim\text{Unif}(A)$ independent.

  4. Define $$S=\bigcup_{A\in \mathcal{A}}\bigcup_{j\le Y_A}\{U_{A,j}\}$$

$S$ is a Poisson process of intensity $\lambda$.

All I already knew was the definition given in the wikipedia page

Are these two different or have connection? Can some one help understanding this?

share|cite|improve this question
Sounds like a spatial Poisson process rather than a temporal Poisson process. In the latter, the number of arrivals in an interval $(t_1, t_2]$ of length $\tau = t_2-t_1$ is Poisson$(\lambda\tau)$ where $\lambda$ is the arrival rate; independent for non-overlapping Here we have the number of points in a set $A \in \mathbb R^d$ is Poisson$(\lambda l(A))$ where $l(A)$ is the measure of $A$; independent for disjoint subsets, and so on. – Dilip Sarwate Dec 17 '11 at 2:58
Where did you find this definition? Are you quoting everything verbatim? – Michael Hardy Dec 17 '11 at 3:03
And given the total number of points in $A$, the individual points are independent and uniformly distributed on $A$ which is analogous to a similar property enjoyed by temporal Poisson processes. Note that this does not depend on the value of $\lambda$. – Dilip Sarwate Dec 17 '11 at 3:04

This definition is more general, in that it characterizes both spatial and temporal Poisson processes. Its elements are the following:

  1. A partition $\cal{A}$ of $\mathbb{R}^d$ into measurable sets of finite measure. Carving $\mathbb{R}^d$ into finite-sized zones is done so that the process on the infinite region can be defined as a product of independent processes on finite regions. The actual partition is irrelevant; the same process is defined for any such $\cal{A}$.
  2. Independent random variables $Y_A ∼ \text{Poisson}(\lambda l(A))$ for each $A \in \cal{A}$. The variable $Y_A$ is the number of events occurring in region $A$.
  3. Independent random variables $U_{A,j} ∼ \text{Uniform}(A)$ for each $A \in \cal{A}$ and $j\in\{1,2,3,...\}$. The variable $U_{A,j}$ is the position of the $j$-th event in region $A$, if there is one.
  4. The (a.s. infinite) set of points $S=\bigcup_{A\in \cal{A}}\bigcup_{j\le Y_A}\left\{U_{A,j}\right\}$. Within each region $A$, the number of events is Poisson-distributed; and given the number of events, the locations of the events are uniformly distributed.

Taking $d=1$ and interpreting $\mathbb{R}$ as time gives a standard temporal Poisson process.

share|cite|improve this answer

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.