Lesson 7 of 8 · 18 min

The exponential distribution

Use the probability density function λe−λt\lambda e^{-\lambda t} to model waiting times, derive its cumulative distribution by integration, and find its mean and standard deviation of 1λ\frac{1}{\lambda}.

What you will be able to do

  • Recognise an exponential random variable TT with pdf f(t)=λe−λtf(t) = \lambda e^{-\lambda t}, t≥0t \geq 0, and interpret λ\lambda as a rate.
  • Show that ff is a valid pdf and derive P(T≤t)=1−e−λtP(T \leq t) = 1 - e^{-\lambda t}.
  • Use, and derive with integration by parts, the mean 1λ\frac{1}{\lambda} and standard deviation 1λ\frac{1}{\lambda}.

Calls to a help line arrive at random, at an average rate of 15 an hour. You have just hung up. How long until the next call? It could be ten seconds or twenty minutes; short waits are common and long waits are rare.

When events happen independently and at a constant average rate, the waiting time TT until the next one is an exponential random variable. It also models the lifetime of components that fail suddenly rather than wearing out, and the time until a radioactive nucleus decays.

Interactive

The shape of the exponential pdf

The horizontal axis is the waiting time tt. Increase λ\lambda: the curve starts higher and falls faster, but the total area under it stays 11.

Definition

The exponential random variable

A continuous random variable TT is exponential with parameter λ>0\lambda > 0 if its probability density function is

f(t)={λe−λt,t≥00,t<0f(t) = \begin{cases} \lambda e^{-\lambda t}, & t \geq 0 \\ 0, & t < 0 \end{cases}

λ\lambda is the rate: the average number of events per unit of time. For the help line, λ=15\lambda = 15 per hour, or 0.250.25 per minute.

It is a valid pdf: f(t)≥0f(t) \geq 0 for every tt, and the total area is 11. The upper terminal is infinite, so integrate to kk and let k→∞k \to \infty:

∫0∞λe−λt dt=lim⁡k→∞[−e−λt]0k=lim⁡k→∞(1−e−λk)=1\int_0^\infty \lambda e^{-\lambda t}\,dt = \lim_{k \to \infty}\Big[-e^{-\lambda t}\Big]_0^k = \lim_{k \to \infty}\left(1 - e^{-\lambda k}\right) = 1

because e−λk→0e^{-\lambda k} \to 0 as k→∞k \to \infty.

Key idea

Probabilities are areas

Integrating the pdf from 00 to tt gives the cumulative distribution:

P(T≤t)=∫0tλe−λu du=[−e−λu]0t=1−e−λtP(T \leq t) = \int_0^t \lambda e^{-\lambda u}\,du = \Big[-e^{-\lambda u}\Big]_0^t = 1 - e^{-\lambda t}

so P(T>t)=e−λtP(T > t) = e^{-\lambda t} and P(a<T<b)=e−λa−e−λbP(a < T < b) = e^{-\lambda a} - e^{-\lambda b}. These are not in the formula book. For a continuous variable, << and ≤\leq give the same probability.

Worked example

Waiting for the next call

  • Simple familiar
  • Technology free

The time, TT minutes, between calls to a help line is exponential with a mean of 4 minutes.

a) Determine P(T>6)P(T > 6) exactly.

b) Determine P(2<T<6)P(2 < T < 6) exactly.

3 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Find λ\lambda from the mean: λ=1μ\lambda = \frac{1}{\mu}.

  2. P(T>t)=e−λtP(T > t) = e^{-\lambda t}.

  3. P(a<T<b)=e−λa−e−λbP(a < T < b) = e^{-\lambda a} - e^{-\lambda b}.

Watch out for

  • Using λ=4\lambda = 4, which gives e−24e^{-24}, a probability far too small for a wait of 6 minutes when the mean is 4.

The time, TT minutes, between buses at a stop has pdf f(t)=0.08e−0.08tf(t) = 0.08e^{-0.08t} for t≥0t \geq 0. Determine the mean time between buses.

in min

Key idea

Mean and standard deviation

If TT is exponential with parameter λ\lambda, then

E(T)=1λandσ=SD(T)=1λE(T) = \frac{1}{\lambda} \qquad \text{and} \qquad \sigma = \text{SD}(T) = \frac{1}{\lambda}

The mean and the standard deviation are equal. Both are in the formula book, but you may be asked to derive the mean.

Worked example

Show that the mean of an exponential variable is 1λ\frac{1}{\lambda}

  • Complex familiar
  • Technology free

TT is an exponential random variable with parameter λ>0\lambda > 0. Show that E(T)=1λE(T) = \frac{1}{\lambda}.

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Write E(T)=∫0∞tf(t) dtE(T) = \int_0^\infty t f(t)\,dt as a limit.

  2. Integrate by parts with u=tu = t.

  3. Evaluate on [0,k][0, k], then take the limit.

Watch out for

  • Choosing u=e−λtu = e^{-\lambda t}, which makes the new integral harder.

  • Substituting ∞\infty as if it were a number: write the limit.

Deriving the standard deviation

Integrate by parts with u=t2u = t^2, dvdt=λe−λt\frac{dv}{dt} = \lambda e^{-\lambda t}:

E(T2)=∫0∞t2λe−λt dt=[−t2e−λt]0∞+∫0∞2te−λt dt=0+2λE(T)=2λ2E(T^2) = \int_0^\infty t^2\lambda e^{-\lambda t}\,dt = \Big[-t^2e^{-\lambda t}\Big]_0^\infty + \int_0^\infty 2te^{-\lambda t}\,dt = 0 + \frac{2}{\lambda}E(T) = \frac{2}{\lambda^2}

using t2e−λt→0t^2e^{-\lambda t} \to 0 as t→∞t \to \infty. Then

Var(T)=E(T2)−[E(T)]2=2λ2−1λ2=1λ2,σ=1λ\text{Var}(T) = E(T^2) - [E(T)]^2 = \frac{2}{\lambda^2} - \frac{1}{\lambda^2} = \frac{1}{\lambda^2}, \qquad \sigma = \frac{1}{\lambda}

An exponential random variable TT has mean 33. Determine the exact probability that TT lies within one standard deviation of its mean.

Reads asYour answer shows here as maths.

Common mistake

λ is a rate, not the mean

"On average, 4 customers arrive per minute" gives λ=4\lambda = 4 per minute and a mean wait of 14\frac{1}{4} minute. "The mean time between customers is 4 minutes" gives λ=14\lambda = \frac{1}{4}. Read which one you are told, and keep the time units consistent.

Also remember that f(t)f(t) is a density, not a probability. f(0)=λf(0) = \lambda can be bigger than 11; P(T=t)=0P(T = t) = 0 for any single value.

Summary

  • T∼T \sim exponential with rate λ>0\lambda > 0 has pdf f(t)=λe−λtf(t) = \lambda e^{-\lambda t} for t≥0t \geq 0.

  • It is valid: f(t)≥0f(t) \geq 0 and ∫0∞λe−λt dt=1\int_0^\infty \lambda e^{-\lambda t}\,dt = 1 (an improper integral).

  • P(T≤t)=1−e−λtP(T \leq t) = 1 - e^{-\lambda t} and P(T>t)=e−λtP(T > t) = e^{-\lambda t}: derive these, they are not in the formula book.

  • E(T)=SD(T)=1λE(T) = \text{SD}(T) = \frac{1}{\lambda}, from integration by parts.

  • λ\lambda is events per unit time; the mean waiting time is 1λ\frac{1}{\lambda}.

Finished the lesson?

Try the 2 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 7 of 80% read

Save your progress with a free account