134 S1 SECTION 5 — THE NORMAL DISTRIBUTION

Normal Distributions
The normal distribution is everywhere in statistics. Everywhere, I tell you. So learn this well...

The Normal Distribution is ‘Bell-Shaped’

1) Loads of things in real life are most likely to fall ‘somewhere in the middle’,
and are much less likely to take extremely high or extremely low values. f(x)
In this kind of situation, you often get a normal distribution.
2) If you were to draw a graph showing how likely different values are, you’d
end up with a graph that looks a bit like a bell. There’s a peak in the
middle at the mean (or expected value). And the graph is symmetrical —
so values the same distance above and below the mean are equally likely.
mean = m
x

f(x) Area 3) With a graph of a normal distribution, the probability of the random
= P(X " a)
Area variable taking a value between two limits is the area under the graph
Area = = P(X ! b) between those limits. And since the total probability is 1, the
P(a " X " b) total area under the graph must also be 1.
x
a b
f(x)
1
4) These three graphs all show normal distributions with the
same mean (m), but different variances (s2).
Graph 1 has a small variance, and graph 3 has a larger variance
— but the total area under all three curves is the same (= 1).
2

5) The most important normal distribution is the standard normal

3
x distribution, or Z — this has a mean of zero and a variance of 1.
m

Normal Distribution N(m, s2)

• If X is normally distributed with mean m and variance s 2, it’s written X ~ N(m, s 2).
• The standard normal distribution Z has mean 0 and variance 1, i.e. Z ~ N(0, 1).

Use Tables to Work out Probabilities of Z Area is

F (z) = P(Z ! z)
Working out the area under a normal distribution curve is usually hard.
But for Z, there are tables you can use (see p146). You look up a value
of z and these tables (labelled F(z)) tell you the probability that Z ! z. 0 z

Or the probability that Z < z — it’s the same thing.

EXAMPLE Find the probability that:
a) Z < 0.1, b) Z # 0.64, c) Z > 0.23, d) Z $ –0.42, e) Z # –1.94, f) 0.12 < Z # 0.82

Tables only tell you the probability of Z being less than a particular value — use a sketch to work out anything else.
a) P(Z < 0.1) = 0.5398 Just look up z = 0.1 in the tables.
For continuous distributions like Z (i.e. with ‘no gaps’ between
b) P(Z # 0.64) = 0.7389 its possible values): P(Z # 0.64) = P(Z < 0.64). f(z) f(z)

c) P(Z > 0.23) = 1 – P(Z # 0.23) = 1 – 0.5910 = 0.4090

Use the symmetry
=
d) P(Z $ –0.42) = P(Z # 0.42) = 0.6628 of the graph: z z
–0.42 0 0 0.42

e) P(Z # –1.94) = P(Z $ 1.94) = 1 – P(Z < 1.94) = 1 – 0.9738 = 0.0262

f(z) f(z)

f) P(0.12 < Z # 0.82) = P(Z # 0.82) – P(Z # 0.12) Again, draw a graph
= 0.7939 – 0.5478 = 0.2461 and use the symmetry: =
z z
–1.94 0 0 1.94

S1 SECTION 5 — THE NORMAL DISTRIBUTION

 135

The Standard Normal Distribution, Z

You can use that big table of the normal distribution function (F(z)) ‘the other way round’ as well
— by starting with a probability and finding a value for z.

Use Tables to Find a Value of z if You’re Given a Probability

EXAMPLE If P(Z < z) = 0.9554, then what is the value of z?
Using the table for F(z) (the normal cumulative distribution function), z = 1.70 .

All the probabilities in the table of F(z) are greater than 0.5, but you can still use the tables with values less than this.
You’ll most likely need to subtract the probability from 1, and then use a sketch.

EXAMPLE If P(Z < z) = 0.2611, then what is the value of z?

1 Subtract from 1 to get a probability greater than 0.5: 1 – 0.2611 = 0.7389

2 If P(Z < z) = 0.7389, then from the table, z = 0.64 . f(z) f(z)

3 Area = 0.2611 Area = 1 – 0.2611

So if P(Z < z) = 0.2611, then, z = –0.64 . = 0.7389

If P(Z < z) = 0.2611, then z must be negative. z z

0 0 0.64

The Percentage Points Table also Tells You z if You’re Given a Probability
You use the percentage points table in a similar way (you start with a probability, and look up a value for z).
But this time, the probability you start with is the probability that z is greater than a certain number.

EXAMPLE If P(Z > z) = 0.15, then what is the value of z? In the examples above, you started with the
probability that z was less than a particular number.
Using the percentage points table, z = 1.0364 .

You might need to use a bit of imagination and a sketch to get the most out of the percentage points table.

EXAMPLE Find z if: a) P(Z < z) = 0.05, b) P(Z < z) = 0.9, c) P(Z $ z) = 0.01

a) Using the percentage points table, if P(Z > z) = 0.05, then z = 1.6449.
So if P(Z < z) = 0.05, then z must equal –1.6449 . f(z) f(z)

b) If P(Z < z) = 0.9, then P(Z > z) = 0.1.

Area = 0.05
Using the percentage points table, z = 1.2816 . Area = 0.05 z z
0 1.6449 –1.6449 0
c) If P(Z $ z) = 0.01, then z = 2.3263 .
Again, you can treat the “greater than or equals sign“ as a “greater than
sign“, since for a continuous distribution, P(Z # z) = P(Z > z).

The medium of a random variable follows a paranormal distribution...

It’s definitely worth sketching the graph when you’re finding probabilities using a normal distribution — you’re much less
likely to make a daft mistake. So even if the question looks a simple one, draw a quick sketch — it’s probably worth it.

S1 SECTION 5 — THE NORMAL DISTRIBUTION

136

Normal Distributions
Main Headingand Z-Tables
All normally-distributed variables can be transformed to Z — which is a marvellous thing.

Transform to Z by Subtracting m, then dividing by s

1) You can convert any normally-distributed variable to Z by:
i) subtracting the mean, and then This means that if you subtract m from any numbers in the
ii) dividing by the standard deviation. question and then divide by s — you can use your tables for Z.

If X ~ N(m, s 2), then X -

v
n = Z, where Z ~ N(0,1)

2) Once you’ve transformed a variable like this, you can use the Z-tables.

EXAMPLE If X ~ N(5, 16) find: a) P(X < 7), b) P(X > 9), c) P(5 < X < 11)

Subtract m (= 5) from any numbers and divide by s (= 16 = 4) — then you'll have a value for Z ~ N(0, 1).

a) P(X < 7) = P ` Z < 7 - 5j = P(Z < 0.5) = 0.6915 N(5, 16) means the variance is 16 — take the
4 square root to find the standard deviation.

Look up P(Z < 0.5) in the big table on p.146.

f(z)

b) P(X > 9) = P ` Z > 9 - 5j = P(Z > 1) = 1 – P(Z < 1) = 1 – 0.8413 = 0.1587

4

c) P(5 < X < 11) = P ` 5 - 5 < Z < 11 - 5j = P(0 < Z < 1.5)
4 4 z
= P(Z < 1.5) – P(Z < 0) = 0.9332 – 0.5 = 0.4332 0 1.5

Find the area to the left of 1.5 and

subtract the area to the left of 0.

The Z-Distribution Can be Used in Real-Life Situations

EXAMPLE The times taken by a group of people to complete an assault course are normally distributed with a mean
of 600 seconds and a variance of 105 seconds. Find the probability that a randomly selected person took:
a) fewer than 575 seconds, b) more than 620 seconds.

If X represents the time taken in seconds, then X ~ N(600, 105).

It's a normal distribution — so your first thought should be to try and ‘standardise‘ it by converting it to Z.

a) Subtract the mean and divide by the standard deviation: P ^ X < 575h = P c X - 600 1 575 - 600 m
105 105
= P ^ Z 1 -2.44h
So P(Z < –2.44) = 1 – P(Z ! 2.44) = 1 – 0.9927 = 0.0073 .

b) Again, subtract the mean and divide by the standard deviation: P ^ X > 620h = P c X - 600 > 620 - 600 m
105 105
= ^
P Z > 1 . 95h
= 1 - P ^ Z # 1.95h
= 1 - 0.9744 = 0.0256

Transform to Z and use Z-tables — I repeat: transform to Z and use Z-tables...

The basic idea is always the same — transform your normally-distributed variable to Z, and then use the Z-tables.
Statisticians call this the “normal two-step”. Well... some of them probably do, anyway. I admit, this stuff is all a bit weird
and confusing at first. But as always, work through a few examples and it’ll start to click. So get some practice.

S1 SECTION 5 — THE NORMAL DISTRIBUTION

 137

Normal Distributions
Main Headingand Z-Tables
You might be given some probabilities and asked to find m and s. Just use the same old ideas...

Find m and s by First Transforming to Z

EXAMPLE X ~ N(m, 22) and P(X < 23) = 0.9015. Find m.

This is a normal distribution — so your first thought should be to convert it to Z.

X - m 23 - m 23 - m
1 P(X < 23) = P[
2
<
2
] = P[Z <
2
] = 0.9015 Substitute Z for
X-µ
σ .
But 0.9015 isn’t in the percentage points table, so you’ll need to
look it up in the ‘big’ table instead (the one for F(z)).

2 If P(Z < z) = 0.9015, then z = 1.29

23 - m
3 So
2
= 1.29 — now solve this to find m = 23 – (2 " 1.29) = 20.42

EXAMPLE X ~ N(53, s 2) and P(X < 50) = 0.2. Find s.

Again, this is a normal distribution — so you need to use that lovely standardising equation again.

1 P(X < 50) = P[Z < 50 -

v
53 ] = P[Z < – 3 ] = 0.2, and so P[Z > – 3 ] = 0.8
v v
This area
3.
Ideally, you'd look up 0.8 in the percentage points table to find – v is 0.2...
...so this area must
be 0.2 as well.
Unfortunately, it isn't there, so you have to think a bit...

3 ] is 0.8, so from the symmetry of the graph, P[Z > 3 ] must be 0.2 .
2 P[Z > – v v
-3
!
3
!
Z

So look up 0.2 in the 'percentage points' table to find that an area of 0.2 is to the right of z = 0.8416.
3 = 0.8416, or s = 3.56 (to 3 sig. fig.)
This tells you that v

If You Have to Find m and s, You’ll Need to Solve Simultaneous Equations

EXAMPLE The random variable X ~ N(m, s2). If P(X < 9) = 0.5596 and P(X > 14) = 0.0322, then find m and s.

9-µ
1 P(X < 9) = P[Z < σ ] = 0.5596.
9-µ
Using the table for F(z), this tells you that σ = 0.15 , or 9 – m = 0.15s .
14 - µ 14 - µ
2 P(X > 14) = P[Z > σ ] = 0.0322, which means that P[Z < σ ] = 1 – 0.0322 = 0.9678.
14 - µ
Using the table for F(z), this tells you that σ = 1.85 , or 14 – m = 1.85s .

3 Subtract the equations: (14 – m) – (9 – m) = 1.85s – 0.15s, or 5 = 1.7s. This gives s = 5 ÷ 1.7 = 2.94 .
Now use one of the other equations to find m: m = 9 – (0.15 " 2.94) = 8.56 .

The Norman Distribution — came to England in 1066...

It’s always the same — you always need to do the normal two-step of subtracting the mean and dividing by the standard
deviation. Just make sure you don’t use the variance by mistake — remember, in N(m, s 2), the second number always shows
the variance. I’m sure you wouldn’t make a mistake... it’s just that I’m such a worrier and I do so want you to do well.

S1 SECTION 5 — THE NORMAL DISTRIBUTION