2
$\begingroup$

Problem setup:

$X_1=Y_1+Y_0,X_2=Y_2+Y_0$ where $Y_1, Y_2\text{ and }Y_0$ are independent Poisson random variables with parameters $θ_1, θ_2\text{ and }θ_0$, respectively.

I know that the joint probability function of bivariate Poisson distribution is given by:

$P(X_1 = x_1, X_2 = x_2) = e^{-(\theta_{1}+\theta_{2}+\theta_{0})} \displaystyle\frac{\theta_{1}^{x_1}}{x_1!}\frac{\theta_{2}^{x_2}}{x_2!} \sum_{i=0}^{min(x_1,x_2)}\binom{x_1}{i}\binom{x_2}{i}i!\left(\frac{\theta_{0}}{\theta_{1}\theta_{2}}\right)^{i}$

Also, the joint probability generating function is: $exp⁡{(θ_1 (t_1-1)+θ_2(t_2-1)+θ_0(t_1 t_2-1))}$

My question is: how can we derive this probability generating function from the joint probability function?

I tried but I couldn't find this result! If anyone knows, please give me a helping hand.

Also, how can I derive the probability generating function in general for the multivariate case? Thanks in advanced.

$\endgroup$
  • $\begingroup$ In general, I believe you can calculate the generating function $P[X_1]$ from the joint generating function $P[X_1,X_2]$ by noting that $P[X_1] = \cfrac{\partial P[X_1,X_2]}{\partial X_1}$ evaluated at $(X_1,X_2)=(1,1)$. Unfortunately, the added $\theta$ and $Y$ variables make this more convoluted, so I'm not sure if this would help. $\endgroup$ – NoseKnowsAll Aug 17 '15 at 15:36
  • $\begingroup$ I found in the book that the PGF of bivariate Poisson distribution is given by:$exp⁡{(θ_1 (t_1-1)+θ_2(t_2-1)+θ_0(t_1 t_2-1))}$, I asked my teacher and he said to me that I can derive it from the joint probability function (noting that $Y_i$ are independent Poisson random variables with the previous parameters. $\endgroup$ – Lean Bader Aug 17 '15 at 15:43
0
$\begingroup$

Since $X_1=Y_1+Y_0,$ and $X_2=Y_2+Y_0$, the joint pgf $E\left[t_1^{X_1} t_2^{X_2}\right]$ can be neatly found as: $$E\left[t_1^{Y_1+Y_0} t_2^{Y_2+Y_0}\right]$$ The latter does not require the joint pmf of $(X_1,X_2)$ ... just the simple joint pmf of $(Y_0,Y_1,Y_2)$ [recall the $Y_i$ are independent Poisson random variables].

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ That what I tried before. I used this rule but I confused about how can I take $x_i$ from the joint pmf and put instead $y_i$?, $\endgroup$ – Lean Bader Aug 17 '15 at 16:06
  • $\begingroup$ What is the joint pmf of $(Y_0,Y_1,Y_2)$? $\endgroup$ – wolfies Aug 17 '15 at 16:35
  • $\begingroup$ This is my confusing. I don't know! $\endgroup$ – Lean Bader Aug 17 '15 at 16:38
  • $\begingroup$ I begin solving this as follows: $$E\left[t_1^{Y_1+Y_0} t_2^{Y_2+Y_0}\right]=E\left[t_1^{Y_1}t_2^{Y_2}(t_1t_2)^{Y_0}\right]$$, then: $$= t_1^{Y_1}t_2^{Y_2}(t_1t_2)^{Y_0}e^{-(\theta_{1}+\theta_{2}+\theta_{0})} \displaystyle\frac{\theta_{1}^{x_1}}{x_1!}\frac{\theta_{2}^{x_2}}{x_2!} \sum_{i=0}^{min(x_1,x_2)}\binom{x_1}{i}\binom{x_2}{i}i!\left(\frac{\theta_{0}}{\theta_{1}\theta_{2}}\right)^{i}$$, then how can I change $x_i$ or $Y_i$?, I lost here. $\endgroup$ – Lean Bader Aug 17 '15 at 16:47
  • $\begingroup$ Thanks . I finally derive it. $\endgroup$ – Lean Bader Aug 18 '15 at 11:44

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.