Math 311

Class 18

Exponential Distributions

\(\displaystyle f(x) = \begin{cases} 0 & \text{ if } x < 0\\\lambda e^{-\lambda x} & \text{ if } x \ge 0\end{cases}\)

\(\displaystyle F(x) = \begin{cases} 0 & \text{ if } x < 0\\1 - e^{-\lambda x} & \text{ if } x \ge 0\end{cases}\)

\(\displaystyle \operatorname{P}(X > x) = G(x) = 1 - F(x) = \begin{cases} 1 & \text{ if } x < 0\\ e^{-\lambda x} & \text{ if } x \ge 0\end{cases}\)

\(\operatorname{E} X = \frac{1}{\lambda}\)

For each $s , and \(t \ge 0\)

\[\operatorname{P}\left(X > t + s\mid X > t\right) = \operatorname{P}\left(X > s\right)\]

\(r(t) = \lambda\)

Hyperexponential distribution

  • Randomly choosing one of \(n\) exponentially distributed variables with rates \(\lambda_1, \lambda_2, \dots, \lambda_n\), with probabilities \(P_1, P_2, \dots, P_n\).

  • Parameters \(\lambda_1, \lambda_2, \dots, \lambda_n\), \(P_1, P_2, \dots, P_n\), where \(\displaystyle \sum_{j=1}^n P_n = 1\).

  • CDF: \(\displaystyle F(t) = 1 - \sum_{j=1}^n P_je^{-\lambda_j t}\)

  • pdf: \(\displaystyle f(t) = \sum_{j=1}^n \lambda_jP_je^{-\lambda_j t}\)

  • \(\displaystyle r(t) = \frac{\sum_{j=1}^n P_j\lambda_je^{-\lambda_j t}}{\sum_{j-1}^n P_j e^{-\lambda_j t}} \class{fragment}{{}= \sum_{j=1}^n \lambda_j\operatorname{P}\left(T = j\mid X > t\right)} \class{fragment}{\rightarrow \min_j \lambda_j \text{ as } t \rightarrow \infty}\)

Sum of Exponential Variables

Let \(X_1, X_2, \dots, X_n\) are independent indentically distributed exponential random variables with rate \(\lambda\).

\[X_j ~ \operatorname{Exp}(\lambda) \text{ for } j = 1, 2, \dots, n\]

What is the distribution of \(X = X_1 + X_2 + \cdots + X_n\)?

(Waiting time for \(n\)-th occurrence.)

\(X \sim \operatorname{Gamma}(n,\lambda)\), or \(\displaystyle f(t) = \lambda e^{-\lambda t}\frac{(\lambda t)^{n-1}}{(n-1)!}\)

Comparison of Exponential Variables

Let \(X_1\) and \(X_2\) be independent exponentially distributed with rates \(\lambda_1\) and \(\lambda_2\). What is \(\operatorname{P}(X_1 < X_2)\)?

Condition on \(X_1\)

Minimum of Exponential Variables

Suppose \(X_i\) is exponential with rate \(\lambda_i\) for \(i = 1, 2, \ldots, n\) and \(X_i\)’s are independent. Then

\[ \begin{aligned} \operatorname{P}(\min(X_1, X_2, \ldots, X_n) > x) &= \operatorname{P}(X_i > x \text{ for each } i = 1, 2, \ldots, n) \\ &= \prod_{i=1}^n \operatorname{P}(X_i > x)\\ &= \prod_{i=1}^n e^{-\lambda_i} x\\ &= \exp\left(-\left(\sum_{i=1}^n \lambda_i\right) x\right) \end{aligned} \]

In other words, \(\min(X_1, X_2, \ldots, X_n)\) has exponential distribution with the rate \(\lambda = \lambda_1 + \lambda_2 + \cdots + \lambda_n\).

Example:

Suppose you arrive at a post office where there are two clerks that are both busy at the time, but there is no one waiting in line in front of you. At the moment one of the two clerks is free, it will be your turn.

Assume that the service time for clerk \(i\) is exponential with rate \(\lambda_i\), for \(i = 1, 2\). Let \(T\) be the total time you spend at the post office. Find \(\operatorname{E} T\).