In a french book, "Calcul des probabilités" from Foata and Fuchs, I found this theorem, which they call "Poissonization".

"Let $(I_k)_{k \in \mathbb{N}}$ be a sequence of independent variables with Bernoulli distribution, each of the same parameter, $p$. Let $N$ be a variable integer-valued independent from the $I_k$, and define $N_1:= \sum_{k=1\ldots N} I_k$, and $N_2:=\sum_{k=1\ldots N} (1−I_k)$.

Then if $N$ has a Poisson distribution with parameter $\lambda$, then $N_1$ and $N_2$ are independent, and have Poisson laws, of parameters $\lambda p$ and $\lambda (1-p)$.

Conversely, if $N_1$ and $N_2$ are independent, then $N$ has a Poisson distribution."

What does this situation modelize ? Is there a natural problem where this situation arises ? I really don't understand what I am dealing with.


Suppose that $N$ represents the number of customers who have arrived to a store up to time $t$, and that $I_k$ is an indication of whether the $k^\text{th}$ customer is male. Then $N_1$ counts up the number of male customers and $N_2$ counts up the number of female customers.


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.