Stats Yr2 Chapter 3 :: Normal

Distribution
[email protected]
www.drfrostmaths.com
@DrFrostMaths

Last modified: 11th February 2018

Use of DrFrostMaths for practice

Register for free at:

www.drfrostmaths.com/homework
Practise questions by chapter, including past
paper Edexcel questions and extension
questions (e.g. MAT).

Teachers: you can create student accounts

(or students can register themselves).
Chapter Overview

1:: Characteristics of the

2:: Finding probabilities on a standard
Normal Distribution
normal curve.
What shape is it? What “Given that IQ is distributed as , determine
parameters does it have? the probability that a randomly chosen
person has an IQ above 130.”
3:: Finding unknown means/standard
deviations.
In Wales, 30% of people have a height
above 1.6m. Given the mean height is 4:: Binomial Normal Approximations
1.4m and heights are normally How would I approximate using a Normal
distributed, determine the standard distribution? Under what conditions can we
deviation of heights. make such an approximation?

5:: Hypothesis Testing

Teacher Notes: This is a combination of all the
old S1 content combined with aspects of S2
(Normal approximations) and S3! (hypothesis
testing on the mean of a normal distribution)
What does it look like?
The following shows what the probability distribution might look like for a
random variable if is the height of a randomly chosen person.

We expect this ‘bell-curve’ shape, where

we’re most likely to pick someone with a
height around the mean of 180cm, with
𝑓 (𝑥 ) the probability diminishing symmetrically
either side of the mean.

180cm Height in cm ()
A variable with this kind of distribution is
said to have a normal distribution. For normal distributions we tend to draw
the axis at the mean for symmetry.
What does it look like?
𝑓 ( 𝑥)

Compare for example the blue and

red curve. If the standard deviation
is greater, the data is more spread
out, and therefore a randomly
chosen thing is less likely to have a
value near the mean (hence it being
less peaky).

We can set the mean and the standard deviation of the Normal
Distribution. If a random variable is normally distributed, then we write
Normal Distribution Q & A
Q1 For a Normal Distribution to be used, the
variable has to be:
?
continuous

Q2 With a discrete variable, all the probabilities

had to add up to 1.
For a continuous variable, similarly:
𝑓 (𝑥 ) the area under the probability graph has
to be 1. ?
Q3 To find , we could:
find the area between these values.
?
Would we ever want to find say?
Q4 Since height is continuous, the probability someone
?
is ‘exactly’ 200cm is infinitesimally small. So not a
‘probability’ in the normal sense.

170cm 180cm 190cm Height in cm ()

Side Notes: You might therefore wonder what the -axis actually is. It is probability density, i.e. “the probability
per unit cm”. This is analogous to frequency density with histograms, where the -value is frequency density area
under the graph gives frequency. We use rather than , to indicate probability density.
Further Facts
The curve has points of inflection one standard
𝑓 (𝑥 ) deviation from the mean, i.e.
(recall from Pure Year 2 that this is where the curve changes
concavity, and where the gradient is not changing)

The Normal Distribution is symmetrical, i.e.

mean = mode = median

𝜇− 𝜎 𝜇+ 𝜎 Height in cm ()

Just For Your InterestTM: The distribution, with a given mean and given standard
deviation , that ‘assumes the least’ (i.e. has the maximum possible ‘entropy’) is… the
Normal Distribution!
Difficult Proof: https://en.wikipedia.org/wiki/Differential_entropy#Maximization_in_the_normal_distribution
Extra Context: This is important in something called Bayesian Statistics. We often have to choose a suitable distribution for the
‘prior’ in the model (i.e. some ‘hidden’ variable). When making inferences based on observed data, we want to assume as little
as possible about any hidden variable, so using a Normal distribution therefore is the most mathematically appropriate choice.
The 68-95-99.7 rule
The 68-95-99.7 rule is a shorthand
used to remember the percentage
of data that is within 1, 2 and 3
standard deviations from the mean
respectively.
You need to memorise this!

!
of data is within one standard
deviation of the mean.
of data is within two standard
deviations of the mean.
of data is within three standard
deviations of the mean.

For practical purposes we consider

all data to lie within
Only one in 1.7 million values fall outside . CERN used a “5
The histogram above is for a quantity which sigma level of significance” to ensure the data suggesting
is approximately normally distributed. existence of the Higgs Boson wasn’t by chance: this is a 1 in 3.5
million chance (if we consider just one tail).
Source: Wikipedia
Examples
[Textbook] The diameters of a rivet produced by a particular machine, mm,
is modelled as . Find:
a)
b)
Fro Tip: Draw a diagram!

a b

8? 7.8 ?8 8.2

8 is the mean, so by the The standard deviation is 0.2, so the

symmetry of the normal data lies within
distribution, 50% of the area
lies above the mean.
Test Your Understanding
IQ (“Intelligence Quotient”) for a given population is, by definition,
distributed using . Find:
a)
b)
Fro Tip: Draw a diagram!

a b

68%
16%
70
?
100 130 100? 115

We know that 95% of data 68% lies within one standard deviation,
lies within 2 standard so there must be 16% at each tail.
deviations of the mean.
Exercise 3A
Pearson Stats/Mechanics Year 2
Pages 40-41
Getting normal values from your calculator
“Use of Technology” Monkey says:
If for some reason you haven’t got a Classwiz by this point, for Kong’s sake, buy one now. [Surreptitiously
pockets banana handed to me by Casio]. The instructions below assume you have a Classwiz.

Just like a cumulative frequency graph gives the running total of

the frequency up to a given value, a cumulative distribution gives
the running total of the probability up to a given value.

IQ is distributed using . Find

(a)
(b)
(c)
(d)
A diagram, while not
a 1. Press MODE. essential, is useful to
2. Choose DISTRIBUTION (option 7) get a sense of whether
our answer is correct.
3. Choose Normal CD (i.e. “Cumulative Distribution”) We know the
4. Since the lower value is effectively , use any value at probability must be
least below the mean (-100000 will do!). Press = above 0.5, so 0.7257
after each value. sounds about right.
5. Put the upper value as 109.
6. Set and
7. You should obtain
No lower bound, 100 109
so use -100000
Getting normal values from your calculator
IQ is distributed using . Find
(a)
(b)
(c)
(d)

b Use 93 for the lower limit and something Fro Note: vs makes no
difference as the
arbitrarily high (e.g. 100000) for the upper limit.
? distribution is continuous.

93 100

c 𝑃 ( 110 < 𝑋 < 120?)= 0.1613( 4 𝑑𝑝)

d From the graph, we can see that we can find the
probability between 80 and 106, and take away
from 1:
?
80 100 106
Froculator Tip: You will need to switch back to normal calculation mode to do the .
Note however that the ANS key will store the last result from the Distribution Mode!
Test Your Understanding
The criteria for joining Mensa is an IQ of at least 131.
Assuming that IQ has the distribution for a population, determine:
a) What percentage of people are eligible to join Mensa.
b) If 30 adults are randomly chosen, the probability that at least 3 of them will be
eligible to join. (Hint: Binomial distribution?)

a ?
b Let be the number of adults among 30 who are eligible to join.
Froculator Note: You
? absolutely can’t use tables in a
formula booklet here. You’ll
need to use a calculator.
Exercise 3B
Pearson Stats/Mechanics Year 2
Pages 40-41
Inverse Normal Distribution
We now know how to use a calculator to value of the variable to obtain a
probability. But we might want to do the reverse: given a probability of being in a
region, how do we find the value of the boundary?

[Textbook] . Find, correct to two decimal b

(i.e. if 40% is above , then 60%
places, the values of such that: must be below)
a.
b. 0.6
?
c.
20 𝑎
1. MODE Distributions
a 2. Choose ‘Inverse Normal’. c We need to find the area up to . From the
3. Put the area as 0.75 (this is the area up sketch we can therefore see we first find
to the value to determine). Put and . the area up to 16 and add on the 0.3:
4. You should get 22.0235.
0.3
?
16𝑎20
The ‘area’ requested by your
calculator is the probability up to the REALLY, DRAW A SKETCH!
value of interest (in this case )

0.75
20 𝑎
DRAW A SKETCH!
Further Example
If the IQ of a population is distributed using .
a. Determine the IQ corresponding to the
top 30% of the population.
b. Determine the interquartile range of IQs.

a b

0.70 0.50
0.25
100 𝑎 𝑄1 100 𝑄3
? ?
𝑃 ( 𝑋>𝑘 ) =0.3 In general the quartiles of a
normal distribution are
approximately
 Test Your Understanding
. Using your calculator,
a. determine the such that
b. determine the such that
c. determine the such that
d. determine the interquartile range of .

a c

𝑃 ( 𝑋>𝑎) =0.65 𝑃?( 𝑋 <76 )=0.2839

0.2839

0.75 ? 0.2

80 𝑎 𝑐76 80

b d

0.4 ? 0.50
𝑃 (?𝑋<𝑄1 )= 0.25∴𝑄1=75.28
0.2375 0.25
75 80 𝑏 𝑄1 80 𝑄3
Exercise 3C
Pearson Stats/Mechanics Year 2
Pages 46-47
Standard Normal Distribution
! is the number of standard deviations above the mean.

If again we use IQ distributed as then: This formula makes sense if you think about
(in your head!) the definition above. For an IQ of 130:
as expected.

IQ Z
100 0 ?  represents the coding:
130 2 ? and . is known as a standard normal distribution.
85 -1 ?
165 4.333
?
62.5 -2.5
?
The 0 and 1 of are the result of the coding. If we’ve subtracted from each value the mean of
the normal distribution is now 0. If we’ve divided all the values by the standard deviation is
now
Standard Normal Distribution
The point of coding in this context is that all different possible normal distributions become a single
unified distribution where we no longer have to worry about the mean and standard deviation.
It means for example when we calculate , this will always give the same probability regardless of the
original distribution.

It also means we can look up probabilities in a -table:

𝜙 ( 𝑎 )=𝑃 ( 𝑍 < 𝑎) ! is the cumulative distribution for the standard

normal distribution. The values of can be found in a -
table.
This is a traditional z-table in the old A This is from the new formula booklet. This is sometimes
Level syllabus (but also found known as a ‘reverse z-table’, because you’re looking up the z-
0 𝑎 elsewhere). You no longer get given this value for a probability. Beware: here it the probability of
and are expected to use your calculator. exceeding rather than being up to . Let’s use it…
Examples
[Textbook] The random variable . Write in terms of for some value of .
(a) (b)

𝑃 ( 𝑋 ≥55 )
a b

? ?
‘Standardise’ to turn into .

[Textbook] The systolic blood pressure of an adult population, mmHg, is modelled as a

normal distribution with mean 127 and standard deviation 16. A medical research wants
to study adults with blood pressures higher than the 95 th percentile. Find the minimum
blood pressure for an adult included in her study.
A sketch will help you
visualise whether you Looking at the table, we see that the top 5%
expect a positive or corresponds to a -value of 1.6449.
negative value of .

? Use formula as soon as

you determine .

5%
0 𝑧
 Further Examples
(a) Determine
(b) Determine
(c) Determine the such that
(d) Determine the such that
a
𝑃 ( 𝑍>−1.3?)=𝑃 ( 𝑍<1.3 )
Fro Tip: Either changing <
to/from > or changing the sign
(+ to/from -) has the effect of

𝑃 (− 2<𝑍<1)=𝑃 (𝑍<1) −𝑃 (𝑍<−2)

“”. However, if you change
b
−1.3 0 1.3 both, the “”s cancel out!

?
𝑃 ( 𝑎<𝑍<𝑏) = 𝑃 ( 𝑍<𝑏 ) − 𝑃 ( 𝑍<𝑎 )
c
From the second sketch we can see that
our value is in the bottom half, so it is
0.7 negative. However, our table only gives d By symmetry, if 0.6 at
positive values of .
the centre, must be 0.2
0 0.5244 From the graphs on the left, we can see
at each tail.
?
by symmetry that the value for the top
30% must be the negation of the -value
?
for the bottom 30% (i.e. top 70%).
0.6
𝑃 ( 𝑍 >𝑎 )=0.7
0.7 0.2 0.2
−𝑎 0 𝑎
−0.5244 0
 Test Your Understanding
1 IQ is distributed with mean 100 and standard deviation 15. Using an
appropriate table, determine the IQ corresponding to the
(a) top 10% of people.
(b) bottom 20% of people.

a
2 If , determine, in terms of :
? (a)
(b)
b a

? ?
b
3 Find the such that:
(a) ?
(b)

a Using appropriate sketch:

?
b
𝑃 ( 𝑍>𝑎?)=0.15
Exercise 3D
Pearson Stats/Mechanics Year 2
Page 49
Missing and
In the last section, you may have thought, “what’s the point of standardising to when I
can just use the DISTRIBUTION mode on my calculator?”
Fair point, but both forward and reverse normal lookups on the calculator required you
to specify and .

[Textbook] . Given that , find the value of .

𝜇20
0.2
𝑃 ( 𝑋>20 )= 0.2
?

The method here is exactly

[Textbook] A machine makes metal sheets with width, cm, modelled the same as before:
as a normal distribution such that . 1. Using a sketch, determine
(a) Given that , find the value of . whether you’re expecting a
(b) Find the 90th percentile of the widths. positive or negative value.
2. Look up value, using

𝑃 ( 𝑋<46?)=0.2119 𝑋 𝑁 ( 50,5 ) 2 tables if you can (otherwise

0.2 your calculator). Make
negative if in bottom half.
3. Use
0.2 To use inverse normal distribution on calculator,
46 50 note that , so use and .
When both are missing
If both and are missing, we end up with simultaneous
equations which we must solve.

Edexcel S1 Jan 2011

“Use of Technology” Monkey says:

Your Classwiz solves simultaneous equations. Look under the EQUATIONS mode.
Test Your Understanding

Edexcel S1 May 2013 (R)

Edexcel S1 Jan 2002

?
?
Exercise 3E
Pearson Stats/Mechanics Year 2
Pages 51-53
Approximating a Binomial Distribution

𝑝(𝑥)

(#successes)

The graph shows the probability function for different Binomial Distributions. Which
one resembles another distribution and what distribution does it resemble?
When is close to 0.5, and is fairly large, it resembles a normal distribution.
?
The results in the distribution being symmetrical. e.g. For a fair coin toss with 10
throws, we’re just as likely to get 1 Head out fo 10 as we are 1 Tail.
 Approximating a Binomial Distribution
If we’re going to use a normal distribution to approximate a Binomial distribution,
it makes sense that we set the mean and standard deviation of the normal
distribution to match that of the original binomial distribution:

𝜇=𝑛𝑝 ?
?
! If is large and close to 0.5, then the binomial distribution can be approximated by
the normal distribution where

Quickfire Questions: We tend to use the letter to represent

the normal distribution approximation
𝑋 𝐵 ( 10,0.2 ) → 𝑌 𝑁 (2 ?, 1.6) of the distribution .

𝑋 𝐵 ( 20,0.5 ) → 𝑌 𝑁 (10 ? ,5) Why use a normal approximation?

• Tables for the binomial distribution only goes up to and your
𝑋 𝐵 ( 6 , 0.3 ) → 𝑌 𝑁 (1.8
? ,1.26 ) •
calculator will reject large values of .
The formula for makes use a factorials. Factorials of large
numbers cannot be computed efficiently. Type in 65! for
example; your calculator will hesitate! Now imagine how
many factorials would be required if you wanted to find . 
Continuity Corrections
One problem is that the outcomes of a binomial distribution (i.e. number of successes)
are discrete whereas the Normal distribution is continuous.
We apply something called a continuity correction to approximate a discrete
distribution using a continuous one.

The random variable represents the time to finish a race in hours. We’re
interested in knowing the probability Alice took 6 hours to the nearest hour.
How would you represent this time on a number line given hours is discrete?
And what about if hours was now considered to be continuous (as )?

Discrete:

Continuous: ?
3 4 5 6 7 8 9 10
We can’t just find when is continuous, because the probability is effectively 0. But would
seem a sensible interval to use because any time between 5.5 and 6.5 would have rounded to 6
hours were it discrete.
 Continuity Corrections
If is a discrete variable, and is its continuous equivalent, how would you represent for
?

Discrete: 𝑋≥5
Continuous: ? 𝑌 ≥ 4.5
Notice the range has been
enlarged by an extra 0.5.
3 4 5 6 7 8 9 10
How would represent for ?

Discrete: 𝑋<9 → 𝑋 ≤ 8
Continuous: ? 𝑌 ≤ 8.5

3 4 5 6 7 8 9 10
! A continuity correction is approximating a
discrete range using a continuous one.
1. If > or <, convert to first.
2. Enlarge the range by 0.5.
Examples

Discrete Continuous
?
?
?
?
?
?
?
?
! A continuity correction is approximating a
discrete range using a continuous one.
1. If > or <, convert to first.
2. Enlarge the range (at each end) by 0.5.
Full Example
[Textbook - Edited] For a particular type of flower bulbs, 55% will produce yellow
flowers. A random sample of 80 bulbs is planted.
(a) Calculate the actual probability that there are exactly 50 flowers.
(b) Use a normal approximation to find a estimate that there are exactly 50 flowers.
(c) Hence determine the percentage error of the normal approximation for 50 flowers.

a
?

b Percentage error:
?

c
?
Test Your Understanding
Edexcel S2 Jan 2004 Q3

?
?

?
Exercise 3F
Pearson Stats/Mechanics Year 2
Page 55
Hypothesis Testing on the Sample Mean
0 1 2 3 4 5 6 7 8 9

Imagine we have 10 children, one of each age between 0 and 9.

This is our population. There is a known population mean of

𝑥 Suppose we took a sample of 4 children.

Sample 1: 1 3 7 8 4.75
Sample mean Tally
Sample 2: 6 2 0 9 4.25
4.00
… 4.25

The mean of this sample is . This sample mean 4.50

is close the true population mean , but is 4.75
naturally going to vary as we consider different 5.00
samples.
Distribution of Sample Means 𝑋
80
For a different sample of 4, we might obtain a 70 I took 1000 In this histogram, 5
different sample mean. What would happen if we 60
50
random samples. actually means
Frequency

took lots of different samples of 4, and found the 40

30
mean of each? How would these means be 20
10
distributed? 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
Mean of Sample of 4 Units
Hypothesis Testing on the Sample Mean
is our distribution across different sample
means as we consider different samples.
Distribution of Sample Means
80 1000 samples.
70
60 Size of each sample: 4 Question 1: What type of distribution is ?
50 children.
Frequency

40 From the left it seems like it is approximately

30
20
10
normally distributed! ?
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 Question 2: On average, what sample mean
Mean of Sample of 4 Units
do we see? (i.e. the mean of the means!)
. The sample means vary around the
Distribution of Sample Means
?
population mean , but on average is .
250
1000 samples. Question 3: Is the variance of (i.e. how
200 Size of each sample:
10 children.
spread out the sample means are) the same
Frequency

150
100 as that of the variance of the population of
50 children?
0 No! On the left, we can see that how spread
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
out the sample means are depends on the
Mean of Sample of 10 Units
sample size. If the sample size is small, the
! For a random sample of size taken from ?
sample means are likely to vary quite a bit.
But with a larger sample size, we expect the
a random variable , the sample mean is different to be closer to the population
normally distributed with mean .
Examples
[Textbook] A certain company sells fruit juice in cartons. The amount of juice in a carton has
a normal distribution with a standard deviation of 3ml.
The company claims that the mean amount of juice per carton, , is 60ml. A trading inspector
has received complaints that the company is overstating the mean amount of juice per
carton and wishes to investigate this complaint. The trading inspector takes a random
sample of 16 cartons and finds that the mean amount of juice per carton is 59.1ml.

Using a 5% level of significance, and stating your hypotheses clearly, test whether or not
there is evidence to uphold this complaint.
Fro Note: Don’t
? As always, state the
distribution and its confuse and . The is

?
Let be the amount of juice per carton. distribution under .
the distribution over
amounts of drink in
Assuming , ? each individual
? Use to work out the
distribution of the
carton.
is the distribution

so there is insufficient ?
sample means, . over sample means,
evidence to reject and conclude that the i.e. the possible
sample means we
mean amount of juice in the whole population is less than 60 see as we take
ml. samples of 16
? cartons. might not
be normally
distributed, but will
be.
Finding the critical region
[Textbook] A machine products bolts of diameter where has a normal distribution with mean 0.580 cm
and standard deviation 0.015 cm. The machine is serviced and after the service a random sample of 50
bolts from the next production run is taken to see if the mean diameter of the bolts has changed from
0.580 cm. The distribution of the diameters of bolts after the service is still normal with a standard
deviation of 0.015 cm.

(a) Find, at the 1% level, the critical region for this test, stating your hypotheses clearly.
The mean diameter of the sample of 50 bolts is calculated to be 0.587 cm.
(b) Comment on this observation in light of the critical region.

a ? This is a two-tailed test so want the

Assuming , ? value corresponding to the top 0.5%
and bottom 0.5% respectively. Use
? table below or your calculator.

?
Use the usual formula , except this
Critical region is or time we use the standard deviation of :
Observed value falls within critical region so sufficient evidence at 1%
level that mean diameter has changed from 0.580cm.
?

b
?
Test Your Understanding
Edexcel S3 June 2011 Q7a

?
Exercise 3G
Pearson Stats/Mechanics Year 2
Pages 58-60
Conditional Probabilities
This is not in the textbook. But given the recent Chapter 2 on Conditional
Probabilities and the fact that the type of question below occurred frequently in S1
papers, it seems worthwhile to cover!

Edexcel S1 May 2014(R) Q4

𝑃 ( 𝑋>𝜇+15|𝑋>𝜇−15)
The intersection of “above ” and
“above ” is just “above ” because the

0.35
? stronger statements takes precedent.
If I said “my age is above 20, and
above 30”, this is equivalent to saying
𝜇−15 𝜇 𝜇+15 just “my age is above 30”.
Test Your Understanding
Edexcel S1 Jan 2013 Q4a,c

?
One Last Toughie…
IQ is distributed as .
A person is considered to be ‘pretty smart’ if their IQ is at least 130.
Determine the median IQ of ‘pretty smart’ people.

Let be the median IQ of ‘pretty smart’ people.

We could have obtained directly at this point

using the calculator’s inverse normal distribution
functionality.