$X_1,...,X_n$ are independent, and $X_i\sim Poisson(i)$, the question ask for the $pdf$ of $\overline X$, and the $pmf$ of $X_{(n)}$.

Here is what I tried:

1. Since they are $Poisson$ distribution, so from $mgf$ of $Poisson$ distribution, $M_{X_{i}}(t)=e^{i(e^t-1)}$. Then since $X_i$ are indep, and let $Y=\sum{X_i}$, so, $M_Y(t)=E(e^{tY})=E(e^{t\sum{X_i}})=E(e^{tX_1} \times \cdots\times e^{t\sum{X_i}})=E(e^{tX_1})\times\cdots\times E(e^{tX_n})=\Pi_{i=1}^n E(e^{tX_i})=\Pi_{i=1}^n e^{i(e^t-1)}=e^{\frac{n(n+1)}{2}(e^t-1)}$

$Y=\sum X_i\sim Poisson(\frac{n(n+1)}{2})$, then $\overline X=\frac{Y}{n}$, so,

$f_{\overline X}(x)=f_{Y}(nx)|\frac{d(nx)}{x}|=\frac{e^{-\frac{n(n+1)}{2}}}{(nx)!}(\frac{n(n+1)}{2})^{nx}n$. I'm not sure whether I was wrong or not on this part, because the transformation formula I found for the discrete RV and continuous RV are different, and here, the transformation I used is from a discrete RV to a continuous RV, I'm got a little confused here whether I should add $n$ as the end of the $pdf$, or may be I was totally wrong.

2. The $pmf$ of $X_{(n)}$ is $P(X_{(n)}=x)=P(X_{(n)}\le x)-P(X_{(n)}\le x-1)=P(X_{1}\le x,\cdots,X_{n}\le x)-P(X_{1}\le x-1,\cdots,X_{n}\le x-1)=\Pi_{i=1}^n P(X_{i}\le x)-\Pi_{i=1}^n P(X_{i}\le x-1)=\Pi_{i=1}^n\sum_{t=0}^{x}\frac{e^{-i}}{t!}i^t-\Pi_{i=1}^n\sum_{t=0}^{x-1}\frac{e^{-i}}{t!}i^t$.

so here, can I switch the product and the summation? Moreover, through this way, I feel I'm already messed up with the calculation.


the transformation I used is from a discrete RV to a continuous RV...

A random variable is discrete if it only can take on a discrete number of values. Since $\bar{X}$ is a finite sum of discrete random variables so it is also discrete. So you should not have the factor of n. In this case you can see \bar{X} is discrete because it can now take on number m/n where m is an integer (because it is a sum of the values of other Poissonian distributions that are integer valued), these are are still fractions which is a discrete set.

With regards to the pmf of $X_{(n)}$. You cannot switch the sum, for example $(a+b)(c+d) \ne ac +db$ which would be equivalent to switching products and sums. Also you did not define $X_{(n)}$, but I'm assuming from your calculation it is $\max_i X_i$.

  • $\begingroup$ Thank you, I did work it out with your help! Your idea is totally correct! $\endgroup$ – user456649 Jun 26 '17 at 22:13

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.