You are on page 1of 13

16 The Exponential Distribution

Example:
1. You have observed that the number of hits to your web
site follow a Poisson distribution at a rate of 2 per day.

Let T be the time (in days) between hits.

2. You observe the number of calls that arrive each day


over a period of a year, and note that the arrivals follow
a Poisson distribution with an average of 3 per day.

Let T be the waiting time between calls.

3. Records show that job submissions to a busy computer


centre have a Poisson distribution with an average of 4
per minute.

Let T be the time in minutes between submissions.

4. Records indicate that messages arrive to a computer


server following a Poisson distribution at the rate of 6
per hour.

Let T be the time in hours that elapses between mes-


sages.
Probability Density Function Of Waiting Times

Generally the exponential distribution describes waiting


time between Poisson occurrences

Proof:
Let T = time that elapses after a Poisson event.

P (T > t) = probability that no event occurred in the time


interval of length t.

The probability that no Poisson event occurred in the time


interval [0, t]:
P (0, t) = e−λt .
where λ is the average Poisson occurrence rate in a unit
time interval.
So:
P (T > t) = e−λt ,

Hence the CDF is:


F (t) = P (T ≤ t) = 1 − e−λt

and, working backwards, the PDF is


f (t) = F ′ (t) = λe−λt
• The PDF:

f (t) = λe−λt , t > 0

= 0 otherwise

• The CDF:

Z t
P (T ≤ t) = F (t) = λe−λT dT
0

it
−e−λT
h
= T =0

= −e−λt + 1

i.e.
F (t) = 1 − e−λt , t > 0
= 0 otherwise.
Example:

If jobs arrive every 15 seconds on average, λ = 4 per minute,


what is the probability of waiting less than or equal to 30
seconds, i.e .5 min?

P (T ≤ .5).

4
3
dexp(x, 4)

P (X <= .5)
1
0

0.0 0.5 1.0 1.5 2.0

Z .5
P (T ≤ .5) = 4e−4tdt
0

i.5
−e−4t
h
= t=0

= 1 − e−2

= 0.86
From R
pexp(.5,4)
[1] 0.8646647
What is the maximum waiting time between two job sub-
missions with 95% confidence?

We need to find k so that


P (T ≤ k) = .95
This is the quantile function.
qexp(.95, 4)
[1] 0.748933
The probability that there will be .74 min, about 45 sec-
onds, between two job submissions is .95.

Applications of the Exponential Distribution:


1. Time between telephone calls

2. Time between machine breakdowns

3. Time between successive job arrivals at a computing


centre
Example
Accidents occur with a Poisson distribution at an average
of 4 per week. i.e. λ = 4
1. Calculate the probability of more than 5 accidents in
any one week
2. What is the probability that at least two weeks will
elapse between accident?

Solution

1. Poisson:
P (X > 5) = 1 − P (X ≤ 5)
In R
1-ppois(5, 4)
[1] 0.2148696
2. Exponential:
P (Time between occurrences > 2)
Z ∞
= λe−λT dT
2

i∞
−e−λT = e−8 = .00034
h
= T =2
In R
pexp(2, 4)
[1] 0.9996645
1-pexp(2, 4)
[1] 0.0003354626
Density of the Exponential Distribution with
λ = 2, 3, 4 and 6

lambda = 2 lambda = 3

2.0

3.0
dexp(x, 2)

dexp(x, 3)

1.5
1.0

0.0
0.0

0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.0

x x

lambda = 4 lambda = 6
4

6
dexp(x, 4)

dexp(x, 6)

4
2

2
0

0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.0

x x

par(mfrow = c(2,2))

curve(dexp(x, 2), 0, 3, main ="lambda = 2")


curve(dexp(x, 3), 0, 3, main ="lambda = 3")
curve(dexp(x, 4), 0, 3, main ="lambda = 4")
curve(dexp(x, 6), 0, 3, main ="lambda = 6")
The Markov Property of Exponential

Examples:

1. The distribution of the remaining life does not depend


on how long the component has been operating.

i.e.
the component does not ‘age’ - its breakdown is a re-
sult of some sudden failure not a gradual deterioration

2. Time between telephone calls

Waiting time for a call is independent of how long we


have been waiting
The Markov property
Show:
P (T ≤ x + t|T > t) = P (T ≤ x)
Proof:
E1 = T ≤ x + t, and E2 = T > t
Then:
P (E1 ∩ E2)
P (E1|E2 ) =
P (E2 )
Now
P (E1 ∩ E2) = P (t < T ≤ x + t)
Z x+t
= λe−λT dT
t

= e−λt [1 − e−λx ]
and Z ∞
P (E2 ) = λe−λT dT = e−λt
t
thus
e−λt [1 − e−λx ]
P (E1 |E2) = −λt
= 1 − e−λx
e
now
1 − e−λx = F (x) = P (T ≤ x)
Examples
1. Calls arrive at an average rate of 12 per hour. Find the
probability that a call will occur in the next 5 minutes
given that you have already waited 10 minutes for a call
i.e.
Find P (T ≤ 15|T > 10)

From the Markov property


P (T ≤ 15|T > 10) = P (T ≤ 5)
So:
P (T ≤ 5) = 1 − e−5λ

The average rate of telephone calls is λ = ·2 in a


minute, then
P (T ≤ 5) = 1 − e−(5)(·2)
= 1 − e−1 = 1 − 0.37 = .63

In R

> pexp(5, .2)


[1] 0.6321206
Examples
2. The average rate of job submissions in a busy computer
centre is 4 per minute. If it can be assumed that the
number of submissions per minute interval is Poisson
distributed, calculate the probability that:
(a) at least 15 seconds will elapse between any two jobs.
(b) less than 1 minutes will elapse between jobs.
(c) If no jobs have arrived in the last 30 seconds, what
is the probability that a job will arrive in the next
15 seconds?
Solution

λ = 4 per minute
(a)
P (t > 15 sec .) = P (T > .25 min)
Z ∞
= λe−λT dT
.25

i∞
−e−λT
h
= T =.25
−1
= e = .37

In R
> 1- pexp(.25, 4)
[1] 0.3678794
Mean of the Exponential Distribution

Recall that when X is continuous:


Z
E(X) = xf (x)dx
x

For the exponential distribution:


Z ∞
E(T ) = tλe−λt dt
0

Integration by parts:
Z ∞ Z ∞
udv = (uv)∞
0 − vdu
0 0

tλe−λt dt = udv

Trick is to spot the u and v:

Take u=t

and dv = λe−λt dt

which gives v = −e−λt

Then
Z ∞ ∞ Z ∞
−λt
−te−λt 0 e−λt dt

tλe dt = +
0 0

∞
−te−λt − λ−1 e−λt

= 0
1
=
λ
R Functions for the Exponential Distribution

• Density Function: dexp


dexp(1, 4)

• Cumulative Distribution Function pexp


pexp(5, 4)
P (T ≤ 5) with λ = 4

• Quantile Function qexp


qexp(.95, 4)
Choose k so that
P (T ≤ k) ≥ .95.
> qexp(.95, 4)
1] 0.748933

You might also like