I saw this problem in Durrett.

Consider two independent Poisson processes $N_1(t)$ and $N_2(t)$ with rates 1 and 2. What is the probability that the two-dimensional process $(N_1(t),N_2(t))$ ever visits the point $(i, j)$?

I wrote this as $P(N_1(t)=i,N_2(t)=j)=P(N_1(t)=i)P(N_2(t)=j)=\frac{e^{-\lambda_1t}(\lambda_1t)^i}{i!}\frac{e^{-\lambda_2t}(\lambda_2t)^j}{j!}$

Is this all that is there to the problem or is there something I am missing?


The process $(N_1(t),N_2(t))_{t\geqslant0}$ starts from $(0,0)$ and jumps from $(i,j)$ to $(i+1,j)$ or to $(i,j+1)$ with respective probabilities $p=\lambda_1/(\lambda_1+\lambda_2)$ and $q=1-p=\lambda_2/(\lambda_1+\lambda_2)$.

Let $h(i,j)$ denote the probability to visit $(i,j)$, then $h(0,0)=1$ and, for every $i\geqslant1$ and $j\geqslant1$, $h(i,0)=p^i$, $h(0,j)=q^j$ and $h(i,j)=ph(i-1,j)+qh(i,j-1)$.

Thus, the reduced variables $c(i,j)=h(i,j)/(p^iq^j)$ are such that $c(i,0)=c(0,j)=1$ and $c(i,j)=c(i-1,j)+c(i,j-1)$ hence one can show recursively that $c(i,j)=(i+j)!/(i!\,j!)$. Finally, for every $(i,j)$, $$ h(i,j)=\frac{(i+j)!}{i!\,j!}\,p^iq^j=\frac{(i+j)!}{i!\,j!}\,\frac{\lambda_1^i\,\lambda_2^j}{(\lambda_1+\lambda_2)^{i+j}}. $$ Sanity check: For every $n\geqslant0$, $\sum\limits_{i=0}^nh(i,n-i)=1$. Do you see why?


I never would thought this as a Markov chain. Nice thought! I get the same answer from a different approach. The joint process must reach i+j total points sometimes. The only concern then is when the joint process reach i+j points, there shall be i points from N1 and j points from N2. So it reduces down to a very simple binomial and that's the answer.


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.