The Normal Approximation to the

Poisson Distribution

Prerequisites
You should be familiar with the normal distribution as an approximation to the binomial
distribution and the use of the continuity correction. You should also be familiar with using
linear combinations (multiples) of a Poisson distribution and with the use of the Poisson
distribution to approximate a binomial distribution.

Example (1)
At a garage the probability selling a one-litre bottle of distilled water in any 1 hour is 0.4.
The garage is open 40 hours a week and 8 hours a day.
(a) (i) Using a binomial distribution find the probability that the garage will sell
exactly 16 one-litre bottles in one randomly chosen week.
(ii) Use a distributional approximation to find the probability that the garage
will sell more than 90 one-litre bottles of distilled water during a
randomly chosen 200 hour period of opening time.
At another garage the probability of selling a one-litre bottle in any 1 hour is 0.05. This
garage is open 60 hours a week. Use a Poisson approximation to find
(b) (i) The probability that the garage will sell exactly 4 one-litre bottles in one
randomly chosen week,
(ii) The probability the garage will sell more than 16 one-litre bottles in one
randomly chosen 5-week period.
Solution
(a ) (i ) Let X denote the number of one litre bottles sold in one week.

X  B  40, 0.4 

 40 
P  X  16      0.4   0.6   0.1279...  0.128 3 s.f.
16 24

 16 
 

(ii ) X  B 200, 0.4  n  200 p  0.4

n  30 p  0.5

© blacksacademy.net

1
 Therefore we may approximate by X  N np ,npq   N  80, 48 

By the continuity correction P  X  90  under X  B 200, 0.4 

corresponds to P  X  90.5 under X  N 80, 48 

x  90.5  80
x  100.5 z   1.516  1.516   0.9352
 48
P  X  90   P  Z  1.516   1  0.9352  0.0648  6.48 % 3 s.f.

(c ) (i ) Let Y denote the number of one litre bottles sold in one week.

Y  B 60, 0.05 n  60 p  0.05

  np  60  0.05  3  5 n  50

Therefore, we may approximate by Y  Po 3

3
4

P Y  4   e 3  0.1680...  0.168 3 s.f.

4!
(ii ) Over 1 week we have Y  Po 3

Therefore, over 5 weeks 5Y  Po 15

P Y  16   1  P Y  16   1  0.6641  0.3359  0.336  3 s.f.

In this question we see the utility of approximating the binomial distribution firstly by the normal
distribution when finding cumulative probabilities for large n, and then by the Poisson
distribution when the probability of success is small and also when n is large. In the second case
the low probability of success (in the example p  0.05 ) renders the direct use of the normal

distribution unsound as an approximation. To use the normal distribution as an approximation

to the binomial distribution we require that p  0.5 .

There remains one missing link in this chain of approximations. Suppose in example (1) we had
gone on to ask the question about a probability over an even larger period than 5 weeks. Say, for
example, we are interested in the likelihood as to whether the garage will sell more than 150 one
litre bottles over a 52 week period. In this case, when we use the Poisson distribution we
encounter a practical difficulty.

Example (1) continued

Using the Poisson approximation found in part (b) of the example (1) above investigate
the probability the garage will sell more than 150 one-litre bottles in one randomly
chosen 52-week period. Find the Poisson approximation suitable for this problem and
explain what practical difficulty you encounter.

© blacksacademy.net

2
 Solution
As before over a 1-week period Y  Po 3 . Therefore over a 52-week period

52Y  Po 3  52  Po 156  . We require P 52Y  150  . The practical problem is that most

tables do not have entries for Poisson parameter   156 . We would also hardly wish to
overcome this practical difficulty by brute force calculation using the definition of the
x
Poisson distribution as P  X  x   e   for x  0, 1, 2, 3, .... by substituting   156 and
x!
summing over values from 0 to 150. Therefore, to deal with this situation we are looking
for another approximation. We wish to be able to replace the Poisson distribution by
another distribution whenever  is large.

The Normal approximation to the Poisson distribution

The result we are looking for is

The normal approximation to the Poisson distribution

Given
1 X  Po   
2  is large    20 
then X may be approximated by X  N   ,  
The continuity correction should be used.

Example (1) continued

Using the Poisson approximation found in part (b) of the example (1) find the probability
the garage will sell more than 150 one-litre bottles in one randomly chosen 52-week
period.

Solution
Over a 1-week period Y  Po 3 .

Over a 52-week period 52Y  Po 3  52  Po 156  .

We require P 52Y  150  .

Since   20, we can approximate Y by R  N 156, 156 .

© blacksacademy.net

3
Using the approximation with the continuity correction

P 52Y  150   P  R  150.5

r  150.5  156
r  150.5 z   0.440   0.440   0.6700
 156
P 52Y  150   P  R  150.5  0.670  3 s.f.

Example (2)
The number of visitors to an Internet website in a day follows a Poisson distribution. On
average there are 30 visitors in a given day. Use the normal approximation to the
Poisson distribution to find the probability of the website receiving between 25 and 32
visitors inclusive in a given day.

Solution
X  Po 30 

Since   20, we can approximate X by Y  N 30,30 

That is,   30 and   30. We require P 25  X  32

Using the approximation with the continuity correction this is approximated by

P 25  X  32  P 24.5  Y  32.5

y1   24.5  30
y 1  24.5 z1    1.0041...
 30
y2   32.5  30
y 2  32.5 z2    0.4565...
 30

P 25  X  32  P 24.5  Y  32.5

 P  1.0041...  Z  0.4565...
 0.3422  0.1761
 0.5183
 51.8 % 3 s.f.

© blacksacademy.net

4
© blacksacademy.net