Arrival probability of Merged Poisson Process

1.7k Views Asked by At

If we have two Poisson process $A$ and $B$ with arrival rate $\lambda_{1}$ and $\lambda_{2}$. Now I have question for calculating the probability of an arrival e.g., from process A in the case of merged Poisson process. Which of the approach we should use.

Considering binomial distribution, given an arrival, the probability that this arrival belongs to process A is $\frac{\lambda_{1}}{\lambda_{1}+\lambda_{2}}$

Or should we use the traditional Poisson pmf to compute the probability of an arrival from process A, such as using, $$p_{m}=\frac{(\lambda_{1}t)^{m}.e^{-\lambda_{1}t}}{m!}$$ and we can put m=1 to get the probability of an arrival from process A using only $\lambda_{1}$ in the equation

1

There are 1 best solutions below

1
On BEST ANSWER

If we have two Poisson process $A$ and $B$ with arrival rate $\lambda_{1}$ and $\lambda_{2}$. Now I have question for calculating the probability of an arrival e.g., from process A in the case of merged Poisson process. Which of the approach we should use.

If you have two independent Poisson processes, then the probabiliy that any particular arrival in the merged process was generated by process $A$ is indeed $\lambda_1/(\lambda_1+\lambda_2)$ .

A Poisson process consists of point events (arrivals) that occur over an interval at a constant average rate but each independently of any other arrival in the process.

Thus in the merged process you have arrivals that occur at a constant joint rate $(\lambda_1+\lambda_2)$ and each independently generated by one from the two independent Possion processes.


Your aproach seems to be asking the condtional probability that, in a very small interval, we get one arrival from process A and no arrival from process B, given that we get only one arrival from the merged process in that very small interval.   Well...sure.

$$\lim\limits_{\Delta t\to 0^+}\cfrac{~\cfrac{(\lambda_1\Delta t)^1e^{-\lambda_1\Delta t}}{1!}\cdot\cfrac{(\lambda_2\Delta t)^0e^{-\lambda_2\Delta t}~}{0!}}{\cfrac{((\lambda_1+\lambda_2)\Delta t)^1e^{-(\lambda_1+\lambda_2)\Delta t}}{1!}}~=~\dfrac{\lambda_1}{\lambda_1+\lambda_2}$$


[Of course, to use this approach you have to first establish that the count of arrivals in the joint process follows a Poisson distribution with the combined average rate.]