This is a simple question but I'm not sure what to do when we're not summing the distributions from $i=1$ to $i=n$.

Let's assume $X_1, \dots, X_n$ are i.i.d observations sampled from a Poisson distribution with parameter $\lambda > 0$. Let $S = \sum_{i=1}^{n} X_i$ and $W = \sum_{i=2}^{n} X_i$.

What are the marginal distributions of $S$ and $W$?

I think the marginal distribution of $S$ is $e^{-\lambda} \sum_{i=1}^{n} \dfrac{\lambda^{x_i}}{{x_i}!}$. But what about W? Do I just take the marginal distribution of $S$ and subtract one Poisson random variable from it?


Since $S$ is the sum of $n$ Poisson random variables with parameter $\lambda$, $S$ should be poisson with parameter $n\lambda$. By similar logic, $W$ is the sum of $n-1$ Poisson random variables with parameter $\lambda$, so $W$ must also be Poisson with parameter $(n-1)\lambda$.

  • $\begingroup$ Ah right, so it's the parameter that changes, not the form of the distribution? Ok, that makes sense, thanks. I'll just wait for a few more upvotes for you and then I'll accept it. $\endgroup$ – meenaparam Jul 27 '18 at 17:32

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.