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

$X_1$ and $X_2$ are discrete stochastic variables. They can both be modeled by a Poisson process with arrival rates $\lambda_1$ and $\lambda_2$ respectively.

$X_1$ and $X_2$ have a constant correlation $\rho$.

Is there an analytic equation that describes the probability density function:

$P(X_1= i,X_2= k)$

share|cite|improve this question
Michael, the simplest possibility was indeed what I had in mind. Thanks for the input learner. I have upvoted both and found this helpful after reading the answers Bivariate Poisson and Diagonal Inflated Bivariate Poisson Regression Models in R – Atlas Nov 26 '12 at 15:55
up vote 2 down vote accepted

Consider this model that could generate correlated Poisson variables. Let $Y$, $Y_1$ and $Y_2$ be three independent Poisson variable with parameters $r$, $\lambda_1$ and $\lambda_2$. Let $$X_i=Y_i+Y$$ for $i=1,2$. Then $X_1$ and $X_2$ are both Poisson with parameters $\lambda_1$ and $\lambda_2$. They have the correlation $$\rho=\frac{r}{\sqrt{(\lambda_1+r)(\lambda_2+r)}}$$ Now the joint distribution can be derived as $$P[X_1=i,X_2=j]=e^{-(r+\lambda_1+\lambda_2)}\sum_{k=0}^{i\wedge j}\frac{r^k}{k!}\frac{\lambda_1^{(i-k)}}{(i-k)!}\frac{\lambda_1^{(j-k)}}{(j-k)!}$$ The case for a bivariate Poisson process is immediate from here.

You could look at the Johnson and Kotz book on multivariate discrete distributions for more information (this construction of a bivariate Poisson distribution is not unique).

share|cite|improve this answer

You haven't given enough information to say what their joint distribution is. But maybe the simplest possibility is this: \begin{align} X_1 & = Y_1 + Y_2 \\ X_2 & = Y_2 + Y_3 \end{align} where $Y_1,Y_2,Y_3$ are independent Poisson-distributed random variables with expectations $\mu_1,\mu_2,\mu_3$, and of course $\mu_1+\mu_2=\lambda_1$ and $\mu_2+\mu_3=\lambda_2$. Recall that the variance of a Poisson-distributed random variable is the same as its expected value. Then the correlation is $\rho$ if $$ \operatorname{cov}(X_1,X_1) = \rho\sqrt{\operatorname{var}(X_1)\operatorname{var}(X_2)}=\rho\sqrt{\lambda_1\lambda_2}. $$ The covariance is $$ \operatorname{cov}(X_1,X_1) = \operatorname{cov}(Y_1+Y_2,Y_2+Y_3) = \operatorname{var}(Y_2)=\mu_2. $$ From there you can think about possible values of $\mu_1,\mu_2,\mu_3$ as functions of $\lambda_1,\lambda_2$.

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.