1.

Application of the Normal Distribution

1.1 Transforming Non-Standard Normal Distributions

Any normal Distribution can be transformed into the Standard Normal Distribution by substituting the
items in the distribution with their appropriate Z-score. The Standard Normal Distribution tables can
therefore be used to predict probabilities in any Normal Distribution.

Example

Consider the case of a manufacturer of electric light bulbs. The light bulbs have a mean life of 1000
hours and a standard deviation of 50 hours. The manufacturer wishes to ascertain the proportion of
bulbs with a life more than 1100 hours.

Recall that Z-score is given by:

𝑋−𝜇
𝑧=
𝜎
The Z-score for the bulbs with lives of 1100 hours is therefore:
1100 − 1000
=2
50
From the normal distribution tables, we can see that only about 2.275% of the bulbs last more than
1100 hours. We can also deduce that about 100 - 2.275 = 97.725% will last less than 1100 hours.

Using Microsoft Excel we can see that 97.725% will last less than 1100 hours.

Therefore, bulbs that will last more than 1100hours = 100 – 97.725 = 2.275%
1.2 Predicting an observation from a Z-score

Let us now consider a condition where the value of an observation is predicted from a given Z-score.
Let us consider again the example of the manufacturer who produces electric bulbs with mean life of
1000 hours and standard deviation of 50 hours. Suppose he wishes to guarantee the lives of his bulbs,
and would be satisfied if only 5% of the bulbs violated the guarantee.

Recall again that Z-score is given by:

𝑋−𝜇
𝑧=
𝜎
From the normal distribution tables, the Z-score with an area of 0.05 (5%) is 1.645. As the value of the
observation should be less than the mean, that is, the observation must be 1.645 standard deviations
below the mean, thus assigned a negative sign (-1.645). The guaranteed life of the bulbs would then
be:
𝑋−1000
−1.645 = 50
= 917.75 hours

Using Excel

1.3 The Normal Approximation to a Discrete Probability Distribution

Under certain circumstances the normal distribution can be used as an approximation to the Poisson
distribution.

Example

Suppose the demand for a component has a Poisson distribution and averages 8 per day. It is required
to know the probability that on any randomly chosen day, demand exceeds 10 components.
The figure below shows the part of the histogram representing the Poisson distribution. The normal
distribution that approximates this Poisson distribution passes through the mid-point of the top of

9 10 11 12
10.5
each rectangle.

This distribution is discrete, and the part of the histogram where demand exceeds 10 components is
reflected by the rectangles for demand of 11 + 12 + 13 + ………….

Therefore, the area to the right of x = 10.5 is required.

The mean of the distribution is 8. We recall that the variance of a Poisson distribution is equal to the
mean. In this case our standard deviation = √8. The Z-score will then be given by:
10.5 − 8
= 0.88
2.828
From normal distribution table, 18.94% (1- 0.8106) of the occasions demand would exceed 10
components. If this probability had been calculated using the Poisson distribution, the value would be
𝜇2 𝜇10
18.41% {1 – [𝑒 −𝜇 (1 + 𝜇 + 2!
+ ……+ 10!
)]} which is quite an acceptable approximation. Practically,
normal distribution gives a good approximation to the Poisson distributions with means exceeding 6.
𝜇2 𝜇10
Using Excel to get the value of [𝑒 −𝜇 (1 + 𝜇 + 2!
+ ……+ 10!
)]
1- 0.815885793 = 0.1841 (18.41%)

1.4 The Distribution of the Variate Sum

Many industrial activity is based on assembling or joining components made in various factories.
Suppose part of a process involves joining end-to-end two metal rods manufactured in two different
factories. This process in reality forms a new population: a rod of each type randomly selected and
joined end-to-end. The lengths of the assembled rods can be described by the distribution of variate
sums. The mean of the variate sums is the sum of the means of the populations from which they were
drawn, that is:

𝜇1 + 𝜇2 + 𝜇3 +. . . … . . +𝜇𝑛
For the standard deviation of the variate sums, we take the square root of their variances, that is:

√𝜎12 + 𝜎22 + 𝜎32 +. . . … … + 𝜎𝑛2

If the populations from which the variate sums were formed are independent and normally
distributed, then the distribution of the variate sums will be normally distributed. For that reason, the
normal distribution can be used to make predictions about such populations.

Example

A manufacturer produces two type of metal rods. Type A has a mean length of 3.5 cm and standard
deviation of 0.03 cm. Type B has a mean length of 4.5 cm and standard deviation of 0.04 cm. A rod
of each type is selected, joined end-to-end, and placed in a slot 8.1 cm long. Assuming the length of
the rods are normally and independently distributed, what proportion of assemblies will not fit the
slot.

Mean lengths of the assembled rods = 3.5 + 4.5 = 8

Standard deviation of lengths of the assembled rods = √0.032 + 0.042 = 0.05

Rods with assembled lengths exceeding 8.1 cm will not fit the slot. The Z-score for length exceeding
8.1 cm is:

Z = (8.1 – 8)/0.05 = 2
From the normal table, 2.275% of the assembled rods would not fit the slot.

Using Excel, it can be shown that 97.725% of the assembled rods would fit the slot;

Therefore, the assembled rods that would not fit the slot = 100 – 97.725 = 2.275%

1.5 The Distribution of the Variate Difference

Under normal circumstances, a distribution of the variate difference would be formed by subtracting
the value of an item selected from one population from the value of an item selected from another
population.

The mean of the variate differences would be the difference in the means of the populations from
which they were drawn, that is:
𝜇1 − 𝜇2

The standard deviation of the variate difference is found by taking the square root of the sum of the
variances from which they were formed, that is:

√𝜎12 + 𝜎22

If the population in which the variate difference were formed are independent and normally
distributed, then the distribution of the variate difference will be normally distributed. For that reason,
the normal distribution can be used to make predictions about such populations.

Example

Full boxes of rice have a mean weight of 225 grams, with a standard deviation of 1.5 grams. It is known
that empty boxes have a mean weight of 7 grams, with a standard deviation of 0.05 grams. The weights
of full and empty boxes are normally and independently distributed. Printed on the box is “Minimum
weight of contents: 214 grams”. What proportion of boxes fail to meet this claim?

Using the distribution of the variate differences, it can be concluded that:

Mean weight of the contents = 225 – 7 = 218 grams

Standard deviation of the weight of the contents = √1.52 + 0.052 = 1.5008 grams

The Z-score:

Z= (214 – 218)/1.5008) = -2.665

From the normal table, only 0.385% of the boxes would fail to meet the claim.

Using Excel we have:

Exercise 1

Packets of tea have a mean weight of 160 grams, standard deviation 2 grams. If the weights are
normally distributed and 500 packets are randomly selected, how many would you expect to weigh

a) Less than 156 grams?

b) Between 158 and 164 grams?
c) At least 162 grams?

Exercise 2

An automatic machine produces packages whose weights are known to be normally distributed with
a standard deviation of 7 grams. To what (mean) weight should the machine be set to ensure that at
least 97.5% of packets have weights in excess of 1600 grams?

Exercise 3

A building contractor erects single storey industrial units to a standard design. The erection process
involves three consecutive tasks, and time taken for each task is:

Task Mean Time Standard

Deviation
Foundation work 10 days 2 days
Brickwork 40 days 5 days
Roofing 20 days 3 days

The firm has won a contract to erect a unit, but will have to pay a penalty of $100,000 if the unit is not
complete within 84 days. The firm can insure against completing after 84 days for a premium of
$5,000.

Would you advise the firm to ensure? (Assume that the times for the tasks are normally and
independently distributed)

Exercise 4

Part of the assembly of a machine involves placing a rod into a slot. Both rods and slots are normally
distributed with the following parameters:
 Mean Standard deviation

Rods 16 cm 0.03 cm

Slots 16.1 cm 0.04 cm

Find the distribution of the clearance between the rods and slots. If a clearance of under 0.2 cm is
satisfactory, find the proportion of the assemblies that fail to meet this requirement.

Reference: Unit 9 (Probability Distributions) of the module IDE –ST 220