Normal Distribution:

Fundamentals
& Using Distributions
Jamie Frost
www.drfrost.org
@DrFrostMaths

Contact the resource team:

resources@drfrost.org
@DrFrostResource

Dr Frost Learning is a registered

Last modified:6th August 2024 charity in England and Wales (no
1194954)
Teacher Notes
Prerequisite
Knowledge Future Links
• Discrete random variables • Hypothesis testing for a
• Standard deviation population mean.
• Central limit theorem.

Throughout the slides, this symbol refers to a web link.

Unless
Key: otherwise specified, this will be to some functionality
within DF.
Key Points Solution step – All slides include
click to reveal pedagogical detail in the
! To be written ‘Notes’ section for each
in books Question/Discussion slide.
Dr Frost Learning is a registered
Prompt charity in England and Wales (no
Using the Dr Frost online platform
TEACHERS STUDENTS
Generate a Start an
random independent
worksheet practice involving
involving skills in skills in this
this PowerPoint PowerPoint.
(for printing or
online task
setting).
drfrost.org/w/73 drfrost.org/p/73
7 7

Clicking this box takes you to a single question practice for a

drfrost.org/s/123a
subskill to allow you further Test Your Understanding opportunities. (e.g.
drfrost.org/s/123a)
Skills in this Lesson
640 Properties of the normal distribution and finding the probability of ranges
640a Know the conditions for the normal distribution to be a suitable model.
640b Understand the notation and use this to sketch normal curves.
640c Calculate the number of standard deviations a value has from the mean.
640d Use the rule to find the probability of values within a given number of standard
deviations of the mean, in context.
640e Use the rule to find the probability of more general ranges, in context.
640f Calculate probabilities with the normal distribution (no context).
640g Dr Frost Learning is a registered
Calculate probabilities with the normal distribution in context.
charity in England and Wales (no
Using the Dr Frost online platform
Clicking this box takes you to a single question practice for a
drfrost.org/s/123a
subskill to allow you further Test Your Understanding opportunities. (e.g.
drfrost.org/s/123a)
Skills in this Lesson
641 Inverse normal distribution
641a Use the inverse normal distribution (no context).
641b Use the inverse normal distribution in context.
641c Determine a quartile of a normal distribution.
641d Use the inverse normal distribution with a double inequality.
641e Use the inverse normal distribution with a double inequality, where the variable is
used twice.
642 Conditional probabilities involving a normal distribution
642a Determine a conditional probability of a range using a normal distribution.
642b Determine the median of a restricted range within a normal distribution.
642c Use the symmetry of a normal distribution to find conditional probabilities.
642d Determine a value at a more general position given a restricted range in a normal
distribution.
642e Solve more difficult problems involving conditional probabilities for a normal
distribution.

Dr Frost Learning is a registered

charity in England and Wales (no
How to use these slides
Though many slides in this resource will have titles specific to the topic, the slide titles in the
table below are used consistently within DFL resources for specific pedagogical purposes.
Any atypical use of a slide type, including any change of animation* or intended use, will be
outlined in the Teacher Notes for the slide.
Slide Title Explanation Default Animations*
To be used as a prior knowledge check or to review
Recap prerequisite knowledge. Can be used as a starter or as part of Green click-to-reveal boxes.
the main lesson.
To be used to highlight key concepts or theorems. This could
Usually in sequence with
The Big include the ‘why’ of the topic - including “real-life” contextual
some green click-to-reveal
Idea scenarios, or putting into context of other mathematical
boxes.
concepts (past and future).
Solution animates in
Example To be modelled by the teacher.
sequence.
Green click-to-reveal boxes.
Test Your
To be completed by students and used for Assessment for For multi-step answers,
Understandi
Learning, primarily using mini-whiteboards. reveal in parts or click final
ng
answer to reveal full solution.
To be used as ‘Example’ &‘Test Your Understanding’ above, Example animates in
To be used as ‘Example’ &‘Test Your Understanding’ above,
Example within the same slide to provide scaffold via visible modelled sequence,
Examplefollowed
animates byinTYU
Example within the same slide to provide scaffold via visible modelled
Problem solution. question with
sequence. Clickgreen click-to-
the header to
Problem solution.
Pair TYU column is blank initially, to focus attention on example. reveal boxes for solution
reveal TYU question, then
Pair TYU column is blank initially, to focus attention on example.
Reveal question by clicking ‘Test Your Understanding’ steps.
green click-to-reveal boxes.
banner.
To be used as fluency practice. Multiple questions in rapid Green click-to-reveal boxes.
Quickfire succession,
To be used for calculations
as fluency that can
practice. be completed
Multiple questions mentally.
in rapid For multi-step
Green answers,
click-to-reveal boxes.
Questions
Quickfire Often used forfor
succession, shorter questions/
calculations that formulae or to isolate
can be completed a small
mentally. reveal in parts oranswers,
For multi-step click final
Questions Often used for shorterpart of the method.
questions/ formulae or to isolate a small line toin
reveal reveal
partsfull solution.
or click final
part of the method. line to reveal full solution.
To be used as a diagnostic question. Multiple choice questions,
Multi-choice with
To be plausible
used distractors,
as a diagnostic to allow
question. teachers
Multiple
Dr Frost to diagnose
choice
Learning is questions, Arrowinpoints
a registered charity Englandto answer,
and Wales on (no
Normal Distribution from Natural
Processes
Consider a black square on a grid.
Suppose that on each move, the square
randomly moves left, right, or
stays where it is.
The square
makes moves
in total.

If we were to repeat this 1. Since each square is equally likely to

with squares and take move left as right, we expect the tallies
to be symmetrical about the centre
tallies of where each square
square.
ends up after moves, what 2. We expect the left and right moves to be
do we expect? roughly balanced, so the final square
positions are likely to be concentrated
around the centre square.
Normal Distribution from Natural
Processes
Final Count
We simulated this Position
random -move
process times and
recorded where each
square ended up.

The distribution
across positions
has a
symmetrical
‘bell-curve’
shape, with the
frequencies
concentrated
at the centre.
Normal Distribution from Natural
Processes
Many real-life I like your
equations.
processes Thanks.Let’s
Let’s
involve such have
havebabies.
babies.

repeated
random
processes.

For example, if people We therefore get the

choose a ‘mate’ with random same distribution, known
height, their choice of as a normal
partner has consequences on distribution.
reducing or increasing the
Concentratio

height of their offspring

relative to them, like the
square moving left and right.
n

Height
Introducing the Normal Distribution

A normal distribution is a
distribution for continuous
variables where the probability
symmetrically tails off from the
mean.
For normal distributions we tend
𝑓 (𝑥 ) to draw the axis at the mean for
symmetry.

cm Height in cm ()
Notation for Normal Distributions
! If a random variable is normally 𝑋 𝑁 ( 𝜇 , 𝜎“with
2
)
distributed, then , where is the mean parameters
and is the standard deviation. and .”
“The random “a normal
variable …” distribution…
”
Why is this peak “has the
lower? distribution
As the data is more 𝑓 (𝑥 ) …”
spread out, a smaller Recall that standard deviation,
roughly speaking, is the average
? of data is
proportion
The distance of values from the mean.
concentrated at the mean is A higher value means the data is
mean. when . more spread out.
The
𝜇=0 ?, 𝜎= 1 mean/
𝜇=0 ?, 𝜎= ? 3 centre is
the
𝜇=− ?2 , 𝜎=? 3 same.
But the
data is
more
spread
𝑥 out.
Test Your Understanding drfrost.org/
s/
640b

1 Note: The reason we

write rather than in is
𝑓 (𝑥 ) because the parameters
A B of ‘’ are the mean and
the variance.
We could write, for
example, to indicate a
variance of , but we’d
typically write as to
make the standard
D
C deviation of explicit.

Match the curve to the distribution.

a𝑋 𝑁 ( 1 ,3 2 ) E? c 𝑋 𝑁 ( 4 ,1 2 ) B
?
b𝑋 𝑁 ( 4 , 22 ) C? d 𝑋 𝑁 ( 1 ,12 ) A?
e 𝑋 𝑁 ( 1 ,2 2) D?
Normal Distribution Q&A
For a normal To find , we would:
distribution to be Find the area?
used, the random between these two
variable has to be... values.
Continuous
?
(e.g. height, Would we ever want to find ?
weight) Since height is continuous, the
With a discrete probability someone is ‘exactly’
𝑓 (𝑥 ) ?
variable, all the cm is infinitesimally small. This is
probabilities had to therefore not a ‘probability’ in
add up to . the normal sense.
For a continuous
variable, similarly: Because of this,
The total area
? under
the probability graph
is .

Heigh
t (cm)
Probability Density
If is meaningless, represents the concentration of probability
what does this (i.e. the probability per unit value), known as
vertical axis probability density. See skill 862.
actually represent? This is analogous to histograms, where the
vertical axis is the frequency density, i.e. the
frequency per unit value.
𝑓 (𝑥 )
Probability
density
The area Frequenc
under a y density
probability
density graph
gives us
probability.

𝑥 𝑥

Analogously, the area under a frequency

density graph (histogram) gives us
frequency.
Test Your Understanding drfrost.org/
s/
640a

2 That data shows some salaries recorded of

adults working in the Republic of Yibistan. a Yibs (to 3sf)
?
Yibs (to 3sf)
Salary ($ Yibs) Freq

b The mean and

Approximat standard deviation of
e shape: the chosen model are
roughly the same as
the mean and
?
standard deviation of
the data.
a Calculate the mean and standard
deviation of salaries in Yibistan. However, the data is
b It is proposed to use the distribution not symmetrical so
to model the salary of a randomly the model is not
chosen person in the population. It is a suitable.
suitable model?

Normal distributions are

not a suitable model
when the data is not
Standard Deviations Above/Below Mean
We are often interested Let and
in how many standard Determine how many
deviations a value is standard deviations
above or below the above or below the mean
mean. This will allow us
𝑓 (𝑥 )
the following values are.
to make calculations 𝜇+ 2 𝜎
about the normal
distribution later.
a 190 standard
deviations
above.
𝜇− 3 𝜎
b 165 standard
deviations
c 21 0 below.
standard
deviations
above.

Height
(cm)
The ‘68-95-99.7 Rule’
! If the data is normally
distributed, then of data is
𝑓 (𝑥 ) within standard deviation
of the mean, i.e.
! and of the data is 𝜇 − 𝜎𝜇+ 𝜎
within and standard
deviations of the mean,
respectively.
𝜇− 2 𝜎 𝜇+ 2 𝜎
We will work out
how to calculate

𝜇− 3 𝜎
𝟔𝟖% 𝜇+ 3 𝜎
these values later,
but it is
worthwhile
remembering
these
percentages
Only one in million values fall 𝟗𝟓 % because they help
outside . CERN used a “ sigma
level of significance” to
us make
judgements
ensure the data suggesting
existence of the Higgs Boson
wasn’t by chance: this is a in
𝟗𝟗.𝟕% mentally about
million chance (if we consider
just one tail).
proportions of
data.
Examples

‘IQ’ is a standardised measure of intelligence,

where the mean IQ is and the standard deviation is
.
Determine the probability that a random selected
person has an IQ:
a
b between and .
between and .

a 𝟎.𝟔𝟖 is one standard b 𝟎.𝟗𝟗𝟕

deviation above .
And one below.
𝑓 (𝑥 ) of values lie within
Noting that
and
one standard , of data is standard
deviation of the deviations of the
mean. mean.

𝜇 − 𝜎 𝜇+ 𝜎
Test Your Understanding drfrost.org/
s/
640d

3 The volume of bottles of juice produced by a factory

have a mean of ml and a standard deviation of ml.
Determine the proportion of bottles that have a
volume between:

a ml and ml.
b ml and ml.

a 𝟗𝟓 %
?
b 𝟔𝟖 %
?
More Complex Ranges
Use a graph when determining more complex ranges using the 68-95-
99.7 rule.
‘IQ’ is a standardised measure of intelligence,
where the mean IQ is and the standard deviation
is.
Determine the probability that a random selected
person has an IQ:
a
b above . If is between
and , then half
between and of this will be

a
𝑓 (𝑥 ) b 𝑓 (𝑥 ) between and .
is within standard
deviations (7). So
by symmetry, the Similarly, if is
remaining is split between and ,
either side. is between
and .
95 %
47.5 %34 %
𝟐.𝟓%
70 130 70 115
𝑃 ( 𝑋 >130 )=𝟎. 𝟎𝟐𝟓 to sf
Test Your Understanding drfrost.org/
s/
640e

4 The diameters of p coins produced by a Royal Mint

can be modelled by in mm.
Determine the probability of:

a
b

a b

? ?
68 % 95 %
16 % 16 % 2.5 % 2.5 %
3.9 44.1 3.844.2

𝑃 ( 𝐷<3.9 ) =𝟎 . 𝟏𝟔
𝑃 ( 𝐷<4.2 )=0.95+0.025
Calculating Arbitrary Ranges
Given that , determine the probability
of:
a b

These are instructions

for the Casio fx-
Use the arrows and Choose Normal CD (CD is used
570/991CW 1 2
OK to select for ranges – you will never use
Distribution. PD).

means Read off the value. Note that if

3 4
so put Lower as and you return to Calculate mode,
Upper as a large value this probability will be stored
(e.g. ). as ANS.
and
Calculating Arbitrary Ranges
Given that , determine the probability
of:
a b

These are
instructions for the
Casio fx-CG50
1 Choose 2 Choose Normal.
DISTRIBUTION.

If , for Tail, choose the Execute, and read off

3 4
third option (F3) to get . the probability .
Enter the values as
shown.
Test Your Understanding drfrost.org/
s/
640f 640g

5 The height in cm of females in the Netherlands can

be modelled using a normal distribution with mean
cm and standard deviation cm.
Use your calculator to determine:

a
b

?
a
b 𝑃 ( 𝐷<165 )=𝟎.𝟐𝟎𝟐?
Inverse Normal Distribution

𝑃 ( 𝑋 < 𝒌 )=0.7
𝑓 (𝑥 )

0.7
𝑘
??
We’ve used the normal distribution to
determine the probability of a range. But
what about the reverse: finding
boundary values that give a specific
probability?
Inverse Normal Distribution on a Calculator
Given that , determine the value of
such that

These are instructions

for the Casio fx- 1 Use the arrows and 2 Choose Inverse Normal.
570/991CW OK to select
Distribution.

The Area is the Read off the value. Note that if

3 4
probability up to the you return to Calculate mode,
value you want to this probability will be stored
determine. Execute. as ANS.
Inverse Normal Distribution on a Calculator
Given that , determine the value of
such that

These are
instructions for the
1 Choose 2 Choose Normal.
Casio fx-CG50 DISTRIBUTION.

To make the in calc, Move the cursor to the

3 4
make the ‘Tail’ . Enter an probability and change
arbitrary value for for to . This will update the
now. Execute. in the range to
Further Examples
Given that , determine the and such that:

a
b

a 𝑃 ( 𝑋 < 𝑎 ) =0.8 b 𝑃 ( 𝑋 < 𝑏 )=0.7

𝒂=𝟏𝟏𝟐. 𝟔 𝒃=𝟏𝟎𝟕 .𝟗

If the area Since is the

above is , the 𝑓 (𝑥 ) mean, of the
𝑓 (𝑥 )
area up to graph will be
is . below it.

0.8 0.2 0.50.2

𝑎
100 𝑏
100
Test Your Understanding drfrost.org/
s/
641a 641b

6 Given that , determine the and such that:

a
b

a
𝑃 (𝑋<
? 𝑎 ) =0.3

𝑃 ( 𝑋 >𝑏 )=0.1
b
?

7 IQ is normally distributed with mean and standard

deviation is . What IQ corresponds to the top of the
population?

𝑃 ( 𝑋>𝑘) =0.01
Let be the IQ for which of
the? population is above.
Harder Examples
Given that , determine the and such that:
a
b

a b

𝑓 (𝑥 ) and are symmetric

about the mean. Via
We can see from
a diagram, we can
the sketch that the
see there’s up to
𝑓 (𝑥 ) area ‘up to ’ is the
area up to , plus
0.8 the area of .
Calculate using
0.1 0.8 0.1 0.5
your calculator.

0.2
100 − 𝑐100
100+ 𝑐 80 𝑑100
Test Your Understanding drfrost.org/
s/
641d 641e

8 Given that , determine the and such that:

a
b

𝑃 ( 𝑋<12 )=0.8413
a

𝑃 ( 𝑋<10+𝑑) =0.7
b

?
Example Test Your
Understanding
We can use the inverse normal 9 The times of athletes in a m
distribution to determine had mean seconds and
quartiles. standard deviation seconds. It
Given that , determine the is suggested the distribution of
upper quartile. times can be modelled using .
Determine the interquartile
range of times according to
this model.
By definition, the
upper quartile is
along the data.

Then use inverse ?

normal as usual.

drfrost.org/ 641c
s/
Conditional Probability with Normal
Distributions
The IQ of a population has the distribution .
A person is defined as ‘smart’ if their IQ is over .
Given a person is smart, determine the
probability that their IQ is over .
Recall that

where ‘|’ means ‘given

that’

𝑓 (𝑥 )

“Above and
above ” is the 100130140
same as just Thinking of ‘given
saying “above ”. that’ as ‘out of’, it’s
The more just the area above
restrictive out of the area
condition ‘wins’. above .
Test Your Understanding drfrost.org/
s/
642a

10 The heights in cm of people in a population

have the distribution .
A theme park ride has a minimum height
restriction of m. Of the people who are able to
ride it, determine the proportion of people with
a height above the mean of the population.

𝑃 ( 𝐻>160 )
𝑃 ( 𝐻>160|𝐻>150?)=
𝑃 ( 𝐻>150 )
Determining the Median of a Restricted
Group
Reflections: We can
The IQ of a population has the distribution . see the calculation of
A person is defined as ‘smart’ if their IQ is over . the two methods are the
same. Method 1 uses
Determine the median IQ of ‘smart’ people. the conditional
probability formula
whereas Method 2
reasons about the
graph.
Method Method
1 Of smart people, 2
will be above 𝑃 ( 𝑋 >130 )= 0.02275
the smart Of this of smart
people median people, they will
() 𝑓 (𝑥 )
be split half and
half above and
below the
median.
0.0114 0.0114

100130𝑎
Use inverse
normal.
Test Your Understanding drfrost.org/
s/
642b

11 The heights in cm of UK adult males can be modelled

using .
The Tall Club only admits people who are at least cm
tall. Determine the median height of people in the tall
club.

cm
Conditional Probabilities Using Symmetry

Given that , determine

Draw a graph to
represent the
information 𝑓 (𝑥 )
given.
This tail is also
by symmetry.

0.3 0.3

𝜇 −10𝜇𝜇+ 10
Test Your Understanding drfrost.org/
s/
642c

12 [Edexcel S1 June 2016 Q6c]

The time, W minutes, taken by women to run a marathon
is modelled by a normal distribution with mean μ minutes.
Given that , find .
(3 marks)

𝑃 ( 𝑊<𝜇−30 )
𝑃 ( 𝑊<𝜇−30|𝑊 <𝜇 )=
? 𝑃 ( 𝑊 <𝜇 )
Value at a More General Position drfrost.org/
s/
642d

We earlier found the median of a restricted group.

We could use the same method to find other positions, e.g. the
lower quartile of a restricted group:

Determine the value of such that

Product lifetime

If a product with limited lifetime keeps functioning for a period of

time, it means the time of failure is after this period.

The lifetime of lightbulbs produced by a factory have a lifetime according to

the distribution in days.
I have two desk lamps in my office. I put a bulb in the first lamp and keep it on
for days, after which it’s still functioning.
I then put a new lightbulb into the second lamp.
Determine the probability that both lightbulbs do not break for the next days.

We want the light bulb

Probability that first bulb lasts:
survives the days in total,
given that it already
survived the first .

The second bulb is new,

so we just want the
Probability that second bulb last: probability it survives it’s
days from the start of its
lifetime.

Probability both last:

Test Your Understanding drfrost.org/
s/
642e

13 [Edexcel A2 June 2018 P3 Q5c Edited]

The lifetime, hours, of a battery has a normal distribution with mean hours
and standard deviation hours. Alice’s calculator requires batteries and will
stop working when any one battery reaches the end of its lifetime. At the
start of her exams Alice put new batteries in her calculator.

She has used her calculator for hours, but has another hours of exams to sit.
Alice only has new batteries so, after the first hours of her exams, although
her calculator is still working, she randomly selects of the batteries from her
calculator and replaces these with the new batteries.
Find the probability that her calculator will not stop working for the remainder
of her exams, giving your answer correct to significant figures.
(5 marks)

At the 16 hour mark, the probability a non- The probability all

new battery lasts at least 20 hours in batteries keep going
total: until the?end of the
exam:
?
The probability a new battery lasts for the
rest of the exam:
Looking Ahead

For all the questions so far, the mean

and standard deviation have been
known. But what if we needed to work
them out?
We’ll explore this in skill 643.
“The heights of people in a population are normally
distributed. of people are taller than m. are shorter than
m.
What is the mean height?”
As we consider different possible
samples, we get different sample
means, which may vary a lot if the
sample size is small. These sample
means form a normal distribution. We
explore this in skill 645.
“The mean IQ of students in a room is thought to be .
A sample of students is taken and the mean IQ is .
Determine the probability that the mean IQ is less than
thought.”
Epilogue : What Makes Normal Distributions So
Special?

P(Athlete 1 Wins)
= 0.99

P(Athlete 2 Wins) Suppose there is a race

= 0.01 between athletes.
What athlete winning
would be more of a
‘surprise’?
Clearly Athlete winning would be more of a
‘surprise’ because it has a lower probability. We say
this probability has more information content or
‘surprisal’, because knowledge that this event
would happen would be more ‘informative’ than
something likely to happen.
Measuring ‘Surprisal’

𝐼 How ‘surprising’ a
probability is can be
calculated using

𝐼 =− log 𝑝
, i.e. an event which is
certain to happen is
‘completely
unsurprising’.

𝑝
1
At the other extreme, ,
so the more unlikely the
event, the more and
more surprising it
becomes.
Entropy of a Distribution

The ‘entropy’ of a distribution is the average

level of information content/surprise across
its outcomes.

Replace each outcome by

its level of ‘surprise’.

Entropy In skill 834, we see that to find the

‘average outcome’ of a random
variable, known as its expected value,
we simply do the weighted average of
the outcomes, each weighted by their
probability.
You can think of higher entropy as a greater
amount of ‘disorder’ to the distribution. You may
be familiar with the concept of entropy from
thermodynamics.
What Distribution has the Greatest
Entropy?

We can calculate entropy for continuous distributions too by

replacing summation with integration: (see skill 864)

𝑓 (𝑥 )
Remarkably, if all we know about a
continuous distribution is its mean
and standard deviation , a normal
distribution has the maximum
entropy across all possible
distributions.
𝜇

We can therefore think of a normal distribution as the

most ‘disordered’ continuous distribution, and is
the most appropriate distribution to pick if we want to
‘assume to least’ about how a random variable
behaves. This is because we want to be as ‘surprised’
as possible on average!
Creating Disorder…
At the start we considered
how represented the single Adding distributions
movement of a square right creates disorder,
(), left () or staying still () because as we add on
more ‘moves’, we can be
less sure where the square
Movement
will end up, i.e. its position
will be more of a ‘surprise’!
We saw that the most
‘disordered’ distribution is a
normal distribution, so the
The final position after
more ’s we add, the more
moves can be found by
normal-like the resulting
adding the outcome
distribution of end-
from copies of this
positions .
distribution:
This would occur regardless
𝐸=𝑀 1+ 𝑀 2+ 𝑀 3 + 𝑀 4 of what the distribution of
was.
This is an informal proof of the Central Limit
Theorem, i.e. the sum of identically distributed
random variables becomes approximately normally
distributed. See skills 850, 868.
𝐸
Use of Normal Distribution in Bayesian
Statistics

𝑃 ( Cause| Effect ¿= 𝑃 ( Effect|Cause ¿ 𝑃 (Cause ) ¿

𝑃 ( Effect )

Bayes’ theorem (skill 636) allows If the causes were discrete and
us to find the probability of categorical, e.g. possible fruits an
object could be, we might use a
various ‘causes’ that leads to an
uniform distribution for the prior
observed ‘effect’, for example distribution.
the classification of a fruit given
observable characteristics like
colour and size.

The prior belief of what different But if the causes was a continuous
causes might be, without value, we would use a normal
considering evidence observed, distribution,
is known as the prior to assume as little as possible.
distribution.