Expected time between events

64 Views Asked by At

Suppose we have two bank tellers.

The time between customer arrivals for the first bank teller is distributed with an exponential distribution with $\lambda_1$. The time between customer arrivals for the second bank teller is distributed with an exponential distribution with $\lambda_2$.

The first teller defers a customer he gets to the supervisor with probability $p_1$ and the second teller defers a customer he gets to the supervisor with probability $p_2$.

In this setup, I am wondering what is the distribution and expected value between customers deferred to the supervisor.

1

There are 1 best solutions below

0
On BEST ANSWER

The number of customers arriving at the first teller in a unit of time, $n_1$, is Poisson distributed:

$$P(n_1=k|\lambda_1)=\frac{\lambda_1^k}{k!e^{\lambda_1}}$$

If the customers are deferred by the first teller to the supervisor independently with a probability $p_1$, then the number of those customers deferred to the supervisor in the same unit time, $m_1$, is distributed according to the binomial distribution:

$$P(m_1=k|n_1,p_1)=\binom{k}{n_1}p_1^k(1-p_1)^{n_1-k}$$

Condition $m_1$ on $\lambda_1$, $p_1$ by marginalizing over $n_1$:

$$P(m_1=k|\lambda_1,p_1)=\sum_{n_1=k}^\infty\frac{n_1!p_1^k(1-p_1)^{n_1-k}\lambda_1^{n_1}}{k!(n_1-k)!n_1!e^{\lambda_1}}$$ $$=\frac{(p_1\lambda_1)^k}{k!e^{\lambda_1}}\sum_{n_1=k}^\infty\frac{(\lambda_1-p_1\lambda_1)^{n_1-k}}{(n_1-k)!}$$ $$=\frac{(p_1\lambda_1)^k}{k!e^{\lambda_1}}e^{\lambda_1-p_1\lambda_1}$$ $$=\frac{(p_1\lambda_1)^k}{k!e^{p_1\lambda_1}}$$ $m_1$ is Poisson distributed with rate $p_1\lambda_1$. Similarly, $m_2$ is Poisson distributed with rate $p_2\lambda_2$, and $m_1+m_2\sim\text{Pois}(p_1\lambda_1+p_2\lambda_2)$. The time between customers being deferred to the supervisor by the two tellers is exponentially distributed with mean $(p_1\lambda_1+p_2\lambda_2)^{-1}$.