Exponential Distribution Cheat Sheet (A Level Only) AQA A Level Further Maths: Statistics

The exponential distribution is a continuous distribution used to model probabilities of waiting time until the first event or intervals between events. It is defined by Mean and Variance
the average number of events within a unit interval and is usually written as 𝑋 ~exp (𝜆). Examples include:
The mean of a continuous probability distribution is given by: Variance is given by:
• Time between phone calls received
∞
• Distance between potholes in a road 𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋)2 ,
𝐸(𝑋) = ∫ 𝑥𝑓(𝑥) ⅆ𝑥.
• Time between collisions of particles in a gas −∞
And 𝐸(𝑋 2 ) is given by:
For an exponential distribution, using the fact that 𝑥 > 0, this becomes:
∞
The conditions required for a situation to be modelled with the exponential distribution are:
∞ 𝐸(𝑋 2 ) = ∫ 𝑥 2 𝜆𝑒 −𝜆𝑥 ⅆ𝑥.
𝐸(𝑋) = ∫ 𝑥𝜆𝑒 −𝜆𝑥 ⅆ𝑥. 0
• Events occur independently 0
• The average rate at which events occur is constant Integrating by parts gives:
• Events occur one at a time The integral can be evaluated by parts:
∞
∞ 𝐸(𝑋 2 ) = −𝑥 2 𝑒 −𝜆𝑥 |∞
0 − ∫ −2𝑥𝑒
−𝜆𝑥
ⅆ𝑥,
In fact, these are the same conditions required by the Poisson distribution. The exponential and Poisson distributions are related in the following way: The Poisson 𝐸(𝑋) = −𝑥𝑒 −𝜆𝑥 |∞
0 − ∫ −𝑒
−𝜆𝑥
ⅆ𝑥 0
0
distribution gives the probability of a number of events occurring within a set interval whereas the exponential distribution gives the probability of an interval This integral has already been evaluated above in the calculation of 𝐸(𝑋).
between events. This relationship is explored mathematically: 1 1
= (0 − 0) − 𝑒 −𝜆𝑥 |∞
0 = 2 2
𝜆 𝜆 𝐸(𝑋 2 ) = (0 − 0) + 𝐸(𝑋) = 2
𝑒 −𝜇 𝜇𝑥
Suppose 𝑋~Po(𝜇) represents the number of busses that arrive in 1 hour so 𝑃(𝑋 = 𝑥) = . Let 𝑌 be the number of busses that arrive within 𝑡 hours such that 𝜆 𝜆
𝑥! 1
𝑒 −𝜇𝑡(𝜇𝑡)𝑦 ⟹ 𝐸(𝑋) =
𝑌~Po(𝜇𝑡), then 𝑃(𝑌 = 𝑦) = and 𝑃(𝑌 = 0) = 𝜇𝑒 −𝜇𝑡 , but 𝑌 = 0 for 𝑡 units of time is the same as 𝑇 > 𝑡, where 𝑇 is the time between intervals. Therefore, 𝜆 Therefore:
𝑦!
𝑇 ~ exp(𝜇).
1 1
This result is expected. If 𝑌~Po(λ), then there are 𝜆 events distributed 𝑉𝑎𝑟(𝑋) = ⟹𝜎=
𝜆2 𝜆
randomly within a unit time. Therefore, it is expected that on average the
Finding Probabilities with the Exponential Distribution 1 1
events will be separated by units of time, i.e. 𝐸(𝑋) = . Just as with the Poisson distribution, the standard deviation and the mean
𝜆 𝜆
If 𝑋~exp (𝜆) then the probability density function is given by: are equal.

𝑓(𝑥) = 𝜆𝑒 −𝜆𝑥 Example 2: Let 𝑁 be the number of beta particles emitted by a radioactive material in a minute. a.) Given 𝑁~ Po(2), write down the distribution of 𝑇, the time
intervals between consecutive emissions. b.) Calculate the probability that the first emission is between 1 and 2 minutes after observation. c.) Write down the
This can be integrated to find the cumulative probability distribution (which gives the probability of an interval between 0 and 𝑥): mean and variance of 𝑇.
𝑥 𝑥
𝑥 a.) Since 𝑁 follows a Poisson distribution, 𝑇 must follow an exponential 𝑇~ exp(2)
𝐹(𝑥) = ∫ 𝑓(𝑥) ⅆ𝑥 = ∫ 𝜆𝑒 −𝜆𝑥 ⅆ𝑥 = −𝑒 −𝜆𝑥 |0 = 1 − 𝑒 −𝜆𝑥
0 0 distribution.
2
b.) Either the probability density function or the cumulative probability
function could be used. Here, the probability density function is used (it is 𝑃(2 > 𝑇 > 1) = ∫ 2𝑒 −2𝑡 ⅆ𝑡
Example 1: Phone calls are received at a consistent, average rate of 3 per hour. Assume that the time intervals between calls are given by an exponential 1
distribution. a.) Calculate the probability of waiting less than 15 minutes for the first call. b.) Calculate the probability of waiting more than 30 minutes for useful to be able to do these calculations in both ways). = −𝑒 −2𝑡 |12
the first call. c.) Given that a call has not been received for 1 hour, what is the probability of waiting at least another 30 minutes for the next call? = −(𝑒 −4 − 𝑒 −2 )
≈ 0.117
c.) The results from above can be used directly. 1
𝐸(𝑇) =
2
a.) The cumulative probability function gives the probability of the interval 𝑇~ exp(3) ⟹ 𝐹(𝑡) = 1 − 𝑒 −3𝑡 1
𝑉𝑎𝑟(𝑇) =
between calls being between 0 and 𝑡 (𝑇/𝑡 is used instead of 𝑋/𝑥 as we are 1 1 3 4
𝑃 (𝑇 < ) = 𝐹 ( ) = 1 − 𝑒 −4
dealing with times). This makes it the easier way to do these types of 4 4
1
problems. Here, 𝑥 = because the set up implies that 𝑥 is in units of hours. ≈ 0.528
4 Example 3: The continuous random variable 𝑋 is modelled by an exponential distribution with mean 𝜇. Find the interquartile range of 𝑋.

b.) More than 30 minutes is the same as not less than 30 minutes. 𝑃(𝑇 > 0.5) = 1 − 𝑃(𝑇 < 0.5)
1 1 1
= 1 − (1 − 𝑒 −2 ) = 𝑒 −2 The distribution of 𝑋 can be written down using 𝐸(𝑋) = 𝜇.
𝑋~ exp ( )
𝜇
≈ 0.607 𝑎 𝑛 𝑎
The 𝑛𝑡ℎ percentile can be found by solving ∫0 𝑓(𝑥) ⅆ𝑥 = for 𝑎. The lower 1 −1𝑥
100 ∫ 𝑒 𝜇 ⅆ𝑥 = 0.25
c.) Here, the conditional probability formula must be used. 𝑃(𝑇 > 1.5) ∩ 𝑃(𝑇 > 1) quartile is found first. 0 𝜇
𝑃(𝑇 > 1.5|𝑇 > 1) = 1 𝑎
𝑃(𝑇 > 1) −𝜇𝑥
−𝑒 | = 0.25
0
1
− 𝑎
But 𝑇 > 1.5 requires 𝑇 > 1.5 so 𝑃(𝑇 > 1.5) ∩ 𝑃(𝑇 > 1) = 𝑃(𝑇 > 1.5) 𝑃(𝑇 > 1.5) 1− 𝑒 𝜇 = 0.25
𝑃(𝑇 > 1.5|𝑇 > 1) = 𝑎
𝑃(𝑇 > 1) = − ln 0.75
3 𝜇
𝑒 −2 1
𝑎 = −𝜇 ln 0.75
= = 𝑒 −2 ≈ 0.607
𝑒 −1 𝑏 𝑏
1 −1 𝑥
The upper quartile is found in the same way, by solving ∫0 𝑓(𝑥) ⅆ𝑥 = 0.75.
∫ 𝑒 𝜇 ⅆ𝑥 = 0.75
0 𝜇
1 𝑏
Parts b and c of the above example have the same answer; because events occur independently, intervals between events must also be independent. This −𝜇 𝑥
−𝑒 | = 0.75
means that information about what has already occurred is irrelevant. The solution could have been found much quicker using this fact. 0
1
− 𝑏
1− 𝑒 𝜇= 0.75
𝑏 = −𝜇 ln 0.25
The interquartile range can now be calculated as the difference between 𝑏 𝑏 − 𝑎 = 𝜇 ln 3
and 𝑎.

https://bit.ly/pmt-cc
https://bit.ly/pmt-edu https://bit.ly/pmt-cc