STA107 – Module 8

Normal Distribution

1
 Continuous Random Variables
• Consider the time it takes for workers in a city to commute to work.

• Let X be the random variable, commute time (in minutes).

• Commute time is a continuous variable.

• Continuous random variables are used to model random variables that can take on any value in an
interval, either finite or infinite.

• Continuous random variables behave similarly to discrete random variables, except that we need to
replace sums of the probability mass function with integrals (calculus needed – coverage in STA256)
of the analogous probability density function.

2
 Special Continuous Distributions
• Uniform distribution

• Normal distribution

• Exponential distribution

• Gamma distribution

• Chi-squared distribution

• Weibull distribution

• Lognormal distribution

• Beta distribution

(You will learn about these distributions in STA256)

3
 Example of A Continuous Random Variable

• A graph of the probability distribution of a

continuous random variable, for example
commute time to work, is a smooth
continuous curve.

• The probability of the interval containing all

the possible values equals 1.

• The area under the curve is exactly 1.

4
 Example of A Continuous Random Variable

What percentage of workers take at least 45

minutes to commute to work?

• About 15% of all workers take more that 45

minutes to commute to work.

• This means that this area equals 15% of the total

area under the curve.

5
 Normal Distribution
• An important class of smooth continuous curves are symmetric unimodal bell-shaped curves known
as normal curves. They describe normal distribution.

• There are other symmetric bell-shaped smooth curves that are not normal (later in the course).

• All normal distributions have the same overall shape.

• Normal distribution is characterized by its mean 𝜇 and its standard deviation 𝜎: 𝑥~ N(𝜇, 𝜎)
• The mean 𝜇 is located at the centre of the symmetric curve and is the same as the median
(and the mode).
• The standard deviation 𝜎 controls the spread of a normal curve.

• Normal distributions are an important model of probability distribution, because they:

1. approximate well the real-world distributions of variables.
2. are most important distribution for statistical inference.
6
 Density Curve and the Normal Model
• To decide how big we expect a data value to be, we need to model the data’s distribution.
• A model will let us say much more precisely how often we’d be likely to see data values of different sizes.

• To model the frequency distribution of a quantitative variable, start by imagining a histogram for a variable based
on lots of data.
• Example:
Histogram of newborns’ weights (in grams)

• A smooth curve may fit on the histogram and

such curves are called density curves.

• Density curve is a curve that is always above the horizontal x-axis.
• It has area exactly 1 underneath it.
• It describes the overall pattern of a distribution.
7
• Aside: The red bars show proportions of newborns with weights between 2500 and 2700 grams.
Two Normal curves with Different Choices for Parameters

8
 Normal Distribution
2
The mathematical definition of the normal distribution begins with the function ℎ 𝑥 = 𝑒 −𝑥 , which produces the bell-shaped
curve shown below, centered at zero and with tails that decay very quickly to zero.

2
By itself, ℎ 𝑥 = 𝑒 −𝑥 is not a distribution since it does not have area 1 underneath the curve. In fact:
This famous result is known as the Gaussian integral.

9
 Normal Distribution
1 −𝑥 2
By rescaling, we arrive at an actual pdf given by 𝑔 𝑥 = 𝑒
𝜋
1
The distribution 𝑔 𝑥 has mean zero and standard deviation ≈ 0.707.
2

1
The inflection points of 𝑔 𝑥 are also at ± and so rescaling by 2 in the direction 𝑥 produces a pdf
2

with standard deviation 1 and inflection points at ±1.

The standard normal random variable Z has probability density function given by

1 2 /2
𝑓 𝑥 = 𝑒 −𝑥
2𝜋

10
 Normal Distribution
By shifting and rescaling Z, we define the normal random variable with mean 𝜇 and standard deviation 𝜎 given by

𝑋 = 𝜎𝑍 + 𝜇
Where we write 𝑋~𝑁 𝜇, 𝜎 .

Some textbooks may write 𝑋~𝑁(𝜇, 𝜎 2 ), so the second parameter in the parenthesis is the variance.
We us the standard deviation, which is consistent with R

Let X be a normal random variable with parameters 𝜇 and 𝜎. The probability mass function of X is given by

𝑥−𝜇
1 −1/2( 𝜎 )2
𝑓 𝑥 = 𝑒 −∞ < 𝑥 < ∞
𝜎 2𝜋

The parameter names are the mean 𝜇 and standard deviation 𝜎.

11
 Translate to
mean 0
𝑋−𝜇
𝑍=
𝜎

Stretch to standard
deviation 1

Notation
Z~N(0,1)
12
 The values 1, 2, 3 in the Empirical Rule are Z-scores
• Z-score, a measure of position, tells us how many standard deviation an observation, x-value,
is away from the mean, 𝜇.

𝑥 − 𝜇 𝑜𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛 − 𝑚𝑒𝑎𝑛
𝑍= =
𝜎 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛

• The Z-score expresses in units of standard deviation.

• No matter what mean or standard deviation 𝑋 has, Z has mean 0, and standard deviation 1.

• If data comes from a normal distribution: x ~ 𝑁 (𝜇, 𝜎)

and therefore, Z-scores have a normal distribution: 𝑍 ~ 𝑁 (𝜇 = 0, 𝜎 = 1)

13
 Standard Normal Distribution
The R function pnorm computes the cdf of the normal distribution, as pnorm(x) = 𝑃 𝑍 ≤ 𝑥 .
Using pnorm, we can compute the probability that Z lies within 1, 2, and 3 standard deviations of its mean

Note:
Since normal distribution is symmetric and bell-shaped,
Empirical Rule holds.

14
 Normal Distribution

Normal Distribution Standard Normal Distribution

X ~ N ( , ) Z ~ N (0,1)
1  ( x −  )2  1  z2 
f ( x) = exp  −  f ( z) = exp  − 
2  2  2
 2  2
E( X ) =  Var ( X ) =  2 E (Z ) = 0 Var ( Z ) = 1

Note –
 any real number
 any positive real
x any real number

X −
Z=  X = Z + 


15
 Normal Distribution
About the Normal distribution:
• Maximum at x = 𝜇
• Inflection points at x = 𝜇 ± 𝜎
• Symmetric about x = 𝜇
• The x-axis is a horizontal asymptote for a normal distribution curve.

16
 Normal Distribution

pdf

CDF

17
Normal Distributions with Fixed Mean and Different SD Values

18
Normal Distributions with Different Mean Values and Fixed SD

19
 Normal Distribution and the Empirical Rule

For any normal random variable, X ~ N(, ),

no matter what  and  are approximately:

• 68% of the normal distribution lies within one standard deviation of the mean.
That is, the area between - and + is about 68%
• 95% lies within two standard deviations of the mean.
That is, the area between -2 and +2 is about 95%
• 99.7% lies within three standard deviations of the mean.
That is, the area between -3 and +3 is about 99.7%. Almost all values fall within 3 standard deviations.

20
For a bell-shaped, symmetric distribution,
the area between - and + is about 68%

21
For a bell-shaped, symmetric distribution,
the area between - and + is about 68%

Z-values are almost 1

in absolute value.

22
For a bell-shaped, symmetric distribution,
the area between -2 and +2 is about 95%

23
For a bell-shaped, symmetric distribution,
the area between -2 and +2 is about 95%

Z-values are almost 2

in absolute value.

24
For a bell-shaped, symmetric distribution,
the area between -3 and +3 is about 99.7%

25
For a bell-shaped, symmetric distribution,
the area between -3 and +3 is about 99.7%

Z-values are almost 3

in absolute value.

26
 For Bell-shaped and Symmetric
Distance from the mean to either quartiles are about 2/3rd of a 𝝈
𝐼𝑄𝑅 4
• A distribution is approx. bell-shaped and symmetric if = ≅ 1.33
𝜎 3

• IQR is roughly (4/3) 𝜎.

27
28
29
30
31
32
For normal distribution,
the plot should be:
• Symmetric about median
• ≈ 0.7% outliers

33
 Table Z (Normal Distribution)
Uploaded on Quercus – see Statistical Tables
Note: The table gives proportion to the left.

34
 Example

A group of Florida State University Psychologists examined the effects of alcohol on the reactions of people
to a threat. After obtaining a specified blood alcohol level, the psychologists placed experimental units in a
room and threatened them with electric shocks. Using sophisticated equipment to monitor the subjects’ eye
movements, the psychologists recorded the startle response (measure in milliseconds) of each subject. The
distribution of startle response was approximately normally distributed with mean 37.9 (milliseconds) and
standard deviation 12.4 (milliseconds).

x ~ N(𝜇, 𝜎)

Startle time ~ N(𝜇 = 37.9, 𝜎 = 12.4)

35
 Example of Proportion Below (Area Below)
What is the probability that experimental subjects had startle response of less than 30 milliseconds?

36
Example of Proportion Below
(Area Below)
What proportion of all experimental subjects had startle response of

less than 30 milliseconds?

Steps:

1. Standardize (find z-score) for

value 30.
𝑥 −𝜇
𝑧=
𝜎

2. Locate z = -0.64 in the z-table.

3. Area below Z of -0.64 is:

37
 Example of Proportion Below (Area Below)
What is the probability that experimental subjects had startle response of less than 30 milliseconds?

38
 Example of Proportion Above (Area Above)
What is the probability that experimental subjects had startle response above 70 milliseconds?

39
 Example of Proportion Above (Area Above)
What is the probability that experimental subjects had startle response above 70 milliseconds?

Steps:

1. Standardize (find z-score) for value 70.

𝑥 −𝜇
𝑍=
𝜎

40
 Example of Proportion Above (Area Above)
What is the probability that experimental subjects had startle response above 70 milliseconds?

P(X > 70) = P(Z > 2.59)

Steps:

2. Locate z-scores of 2.59 in the Z-table.

3. Note Z-table gives area below.

So, we need to do the following:

Area above Z = 2.59 is:

P(Z > 2.59) = 1 – P(Z < 2.59)

41
 Example of Proportion Above (Area Above)
What is the probability that experimental subjects had startle response above 70 milliseconds?

42
 Example of Proportion Between (Area Between)
What is the probability that experimental subjects had startle response of between 40 and 50 milliseconds?

43
 Example of Proportion Between (Area Between)
What is the probability that experimental subjects had startle response of between 40 and 50 milliseconds?

Steps:

1. Standardize (find z-score) for value 40.

𝒙 −𝝁
𝒁=
𝝈

2. Standardize (find z-score) for value 50.

𝒙 −𝝁
𝒁=
𝝈

3. Locate Z-scores of 0.17 and 0.98 in the Z-table.

And, note the area below each:

Area #1: Area below Z = 0.98 is

Area #2: Area below Z = 0.17 is

Subtract Areas: Area #2 – Area#1 =

44
 Example of Proportion Between (Area Between)
What is the probability that experimental subjects had startle response of between 40 and 50 milliseconds?

45
 Example of Proportion Between (Area Between)
What is the probability that experimental subjects had startle response of between 40 and 50 milliseconds?

46
 Example of Proportion Between (Area Between)
What is the probability that experimental subjects had startle response of between 40 and 50 milliseconds?

47
 Example of Proportion Between (Area Between)
Incorrect Way of Calculating
What is the probability that experimental subjects had startle response of between 40 and 50
milliseconds?

Watch out! This is incorrect!

48
 Finding X-value given an Area (a Proportion)
Ten percent of all experimental subjects had startle response (in milliseconds) above what value?

49
 Example of Using Z-table Backward

Ten percent of all experimental subjects had startle response (in milliseconds) above what value?
• Top 10% is equivalent to 0.1000 proportion.
• Search for 0.1000 inside the z-table to find its associated z-score.
• We cannot find the exact 0.1000 proportion but we can find the closest value: 0.1003.
• The associated Z-value to 0.1003 proportion is -1.28.
• However, we need to change the sign to positive sign for z-value,
because we are looking at top 10%.
• So, our Z-score is: Z = +1.28

50
Ten percent of all experimental subjects had startle response above (in milliseconds) above what value?
𝒙 −𝝁
𝒁=
𝝈
Step 1: Find the Z-score that has 0.10 area above it

Step 2: Re-arrange the formula

𝒙 −𝝁 = 𝒛∗𝝈
𝒙 = 𝝁 + (𝒛 ∗ 𝝈)
𝒙=

51
 Example: The 68-95-99.7 Rule for Normal Models

The distribution of heights of women aged 18-24 is approx. normal with mean 64.5 (in inches) and
standard deviation 2.5 (in inches).

• Approximately, the middle 68% of women’s heights is between: __________ and __________ inches.

• Moreover, approximately, 32% (100% - 68% = 32%) of women’s heights fall outside the range of
62 to 67.
• Approx.__________ of women are taller than 67.
• Approx.__________ of women have heights below 62.

52
 Example: The 68-95-99.7 Rule for Normal Models

The distribution of heights of women aged 18-24 is approx. normal with mean 64.5 (in inches) and
standard deviation 2.5 (in inches).

• Approximately, the middle 95% of women’s heights is between: __________ and __________ inches.

• Moreover, approximately, 5% (100% - 95% = 5%) of women’s heights fall outside the range of
59.5 to 69.5.
• Approx.__________ of women are taller than 69.5.
• Approx.__________ of women have heights below 59.5.

53
 Example: The 68-95-99.7 Rule for Normal Models

The distribution of heights of women aged 18-24 is approx. normal with mean 64.5 (in inches) and
standard deviation 2.5 (in inches).

• Approximately, the middle 99.7% of women’s heights is between: __________ and __________ inches.

• Moreover, approximately, 0.3% (100% - 99.7% = 0.3%) of women’s heights fall outside the range of
57 to 72.
• Approx.__________ of women are taller than 72.
• Approx.__________ of women have heights below 57.

54
Suppose a cereal manufacture puts pictures of
Canadian medal winners from the 2016 summer
Olympic Games on cards in boxes of cereals, in the
hope of increasing sales. The manufacture announces
that 20% of the boxes contain a picture of Andre De
Grasse, 30% picture of Christine Sinclair, and the rest
a picture of Penny Oleksiak.

• We will keep the assumption that pictures are distributed at random and will trust the manufactures
claim that 20% of the cards are Andre De Grasses.
• So, when you open the box, the probability that you succeed in finding the De Grasses is 0.2.

55
 Example
Suppose 50 boxes of cereal are randomly selected. What is the probability that we find at most 5
Andre De Grasse pictures?

56
 Normal Approximation to Binomial Distribution

• A Normal model, then, is a close enough approximation only for a large enough number of trials.

• The “large enough” depends on the probability of success.

• We would need a larger sample if the probability of success were very low (or very high).

• Success/Failure Condition:

A Binomial model is approximately Normal if we expect at least 10 success and 10 failures.

𝑛𝑝 ≥ 10 𝑎𝑛𝑑 𝑛 1 − 𝑝 ≥ 10

X ~ Binomial (𝑛, 𝑝)

X ~ Normal (𝜇 = 𝑛𝑝, 𝜎 = 𝑛𝑝(1 − 𝑝))

57
 Normal Approximation to Binomial Distribution

• The usual condition is that 𝑛𝑝 ≥ 5 and 𝑛(1 − 𝑝) ≥ 5

(Note: some sources suggest 5 be replaced by 10 see next slide for explanation)
• A stronger condition is that “the normal fits inside the binomial” or

0 < 𝑛𝑝 − 3 𝑛𝑝(1−p) < 𝑛𝑝 + 3 𝑛𝑝(1−p) < 𝑛

that means, (𝑛𝑝 − 3 𝑛𝑝(1−p), 𝑛𝑝 + 3 𝑛𝑝(1−p)) 𝑖𝑛𝑠𝑖𝑑𝑒 (O, 𝑛)

58
 Why 𝒏𝒑 ≥ 𝟏𝟎 𝒂𝒏𝒅 𝒏 𝟏 − 𝒑 ≥ 𝟏𝟎?
Why 10? (Actually 9)
• Recall that Normal model extends infinitely in both directions.
• But a Binomial model must have between 0 and n successes, so if we use a Normal to approximate an
Binomial, we have to cut off its tails. That’s not very important if the centre of the Normal model is so far from
0 and n that the lost tails have only a negligible area. More than three standard deviation should do it,
because a Normal model has little probability past that.
• So, the mean needs to be at least three standard deviation from 0 and at least three standard deviations from
𝑛.
• Lets look at the 0 end.
We require: 𝜇 − 3𝜎 > 0
Or, in other words: 𝜇 > 3𝜎
For Binomial, that’s: 𝑛𝑝 > 3 𝑛𝑝(1 − 𝑝)
Squaring yields: 𝑛2 𝑝2 > 9𝑛𝑝 1 − 𝑝
Now simplify: 𝑛𝑝 > 9 1 − 𝑝
Since 1 − 𝑝 ≤ 1, we require 𝑛𝑝 > 9
For simplicity, we usually require that 𝑛𝑝 and 𝑛(1 − 𝑝) for the other tail be at least 10 to use Normal
59
Approximation, the success/failure condition.
 Further Note
• Looking at the final step (previous slide), we see that we need 𝑛𝑝 > 9 in the worst case,
when 1 − 𝑝 or 𝑝 is near 1, making the Binomial quite skewed.

• When 1 − 𝑝 or 𝑝 are near 0.5-say between 0.40 and 0.60-the Binomial model is nearly symmetrical
and 𝑛𝑝 > 5 ought to be safe enough.

• Although, we will always check for 10 expected successes and failure, keep in mind that
for values 𝑝 near 0.50, we can be somewhat more forgiving.

60
 Continuity Correction
• The binomial distribution is discrete, and probability is in rectangles.

• When you use a continuous normal distribution to approximate a binomial probability, you need to
move 0.5 unit to the left and right of the midpoint to include all possible x-values in the interval

Exact binomial probability Normal approximation

P(x = c) P(c – 0.5 < x < c + 0.5)

In other words c c–0.5 c+0.5

• Add 0.5 to the value we’re approximating for 𝑃 𝑋 ≤ 𝑥 calculations
• Subtract 0.5 from the value we’re approximating for 𝑃(𝑋 ≥ 𝑥) calculations
For example:
𝑃 𝑋 ≤ 𝑥 = 𝐵𝑖𝑛𝑜𝑚𝑖𝑎𝑙(𝑥; 𝑛, 𝑝) ≈ (area under the normal curve to the left of 𝑥 + 0.5)
𝑥+0.5−𝑛𝑝
= 𝜙( )
𝑛𝑝(1−𝑝) 61
 Normal Approximation to Binomial Distribution
[Link]

Check if Normal approximation to the Binomial distribution is appropriate:

• 𝑛𝑝 = 50(0.20) = 10

• 𝑛(1 − 𝑝) = 50(1 − 0.80) = 40

We can use Normal Approximation to the Binomial.

𝑋 ~ 𝐵𝑖𝑛𝑜𝑚𝑖𝑎𝑙 (𝑛 = 50, 𝑝 = 0.20)

• 𝐸(𝑋) = 𝑛𝑝 = 50(0.20) = 10

• 𝑆𝐷 𝑋 = 𝑛𝑝 1 − 𝑝 = 50 0.20 1 − 0.80 = 2.83

𝑋 ~ 𝑁𝑜𝑟𝑚𝑎𝑙(𝜇 = 10, 𝜎 = 2.83)

62
 Normal Approximation to Binomial

• When we are dealing with a large number of

trials, direct calculation of probabilities become
tedious. (although we can use R to calculate
probabilities).
• Now suppose we open 50 boxes of cereal and
count the number of Andre De Grasse pictures
we find.
• The histogram shows this probability model.
• It is centred at np = 50(0.20) = 10, pictures, as
expected, and it appears to be fairly symmetric
around the centre.
• With this larger sample size, it appears that a
Normal model might be a useful approximation.

63
 Normal Approximation to Binomial
Suppose 50 boxes of cereal are randomly selected. What is the probability that we find at most 5
Andre De Grasse pictures?

𝑋 ~ 𝑁𝑜𝑟𝑚𝑎𝑙(𝜇 = 10, 𝜎 = 2.83)

Without continuity correction

𝑃 𝑋≤5 =

With continuity correction

𝑃 𝑋 ≤ 5.5 =

64
Without continuity correction With continuity correction

65
 Normal Probability Plots (Normal Q-Q Plots; Normal Quantile Plots)

• A histogram can reveal distinctly non-normal features of a distribution.

• If the histogram appears roughly symmetric and unimodal, we use another graph, the normal
quantile plot as a better way of judging the adequacy of a normal model.

66
 Probability Plots
• An investigator will often have obtained a numerical sample consisting of n observations and wish to know
whether it is plausible that this sample came from a population distribution of some particular type.

• Many formal procedures from statistical inference are based on the assumption that the population
distribution is of a specified type.

Caution: The use of such a procedure is inappropriate if the actual underlying probability distribution differs
greatly from the assumed type.

• Understanding the underlying distribution can sometimes give insight into the physical mechanism involved
in generating the data.

• An effective way to check a distributional assumption is to construct a probability plot.

• The basis for our construction is a comparison between percentiles of the sample data and the
corresponding percentiles of the assumed underlying distribution.

67
 Sample Percentiles
• The details involved in constructing probability plots differ a bit from source to source.

• Roughly speaking, sample percentiles are defined in the same way that percentiles of a continuous population
distributions were defined.

• The sample 50th percentile (i.e. the sample median) should separate the smallest 50% of the sample from
the largest 50%.

• The sample 90th percentile should be such that 90% of the sample lies below that value and 10% lies above.

• What if our sample size n was small, for example n = 10?

• We can split off 20% or 30% of the data, but there is no value that will split off exactly 23% of these 10
observations.

• To proceed further, we need an operational definition of sample percentiles (this is one place where different
people and different software packages do slightly different things).

68
 Sample Percentiles
• Statistical convention states that when n is odd, the sample median is the middle value in the ordered list of

sample observations.

• For example, the sixth-largest value when n = 11. This amounts to regarding the middle observation as being

half in the lower half of the data and half in the upper half.

• Similarly, suppose n = 10. we could call the third-smallest value the 25th percentile, we are regarding that

value as being half in the lower group (consisting of two smallest observations).

• This leads to the following general definition of sample percentiles:

Order the n sample observations from smallest to largest. Then the ith-smallest observation in the list is taken to be

the sample [100(i-0.5)/n]th percentile.

69
 Sample Percentiles
Definition: Order the n sample observations from smallest to largest. Then the ith-smallest observation in the list is
taken to be the sample [100(i-0.5)/n]th percentile.

For example, if n = 10, the percentages corresponding to the ordered sample observations are:

• 100(1-0.5)/10 = 5%

• 100(2-0.5)/10 = 15%

• 100(3-0.5)/10 = 25%

• … and, 100(10 – 0.5)/10 = 95%

That is, the smallest observation is designated the sample 5th percentile, the next-smallest value the sample 15th
percentile, and so on.

All other percentiles can then be determined by interpolations; for example, the sample 10th percentile is halfway
between the 5th percentile (smallest sample observation) and the 15th percentile (the second smallest observation)
for n = 10 values.
70
 A Probability Plot
• Let’s consider whether our sample data could plausibly have come from some completely specified population

distribution (e.g., normal distribution with 𝜇 = 10 and 𝜎 = 3).

• If the sample had actually been selected from this distribution, the sample percentiles

(ordered sample observations) should be reasonably close to the corresponding population distribution percentiles.

• That is, 𝑖 = 1, 2, … , 𝑛 there should be reasonable agreement between the ith-smallest sample observation and the

theoretical [100(i-0.5)/n]th percentile for the specified distribution.

71
 A Probability Plot
• Now from the sample percentile pairs, that is, the pairs

(ith-sample observation, [100(i-0.5)/n]th percentile of the population distribution) for i = 1, …, n.

• Each pair can be plotted as a point on a two-dimensional coordinate system.

• If the sample percentiles are close to the corresponding population distribution percentiles, the first number in each

pair will be roughly equal to the second number, and the plotted points will fall close to a 450 line.

• Substantial deviation of the plotted points from a 450 line suggest that the assumed distribution might be wrong.

72
 Normal Probability Plot
A plot of the n pairs

(ith-smallest observation, [100(i-0.5)/n]th z percentile)

on a two-dimensional coordinate system is called a normal probability plot.

If the sample observations are in fact from a normal distribution, then the points should fall close to a
straight line (although not necessarily a 450 line).

Thus, a plot for which the points fall close to some straight line suggests that the assumption of a normal
population is plausible.

73
 Normal Probability Plots (Normal Q-Q Plots; Normal Quantile Plots)

• Normal Probability, or QQ, Plot, is a scatterplot of n ordered pairs, (x, y).

• The horizontal axis is the expected standard normal observations (in order from smallest to largest).

• The vertical axis is the Z-scores of the observations (in order from smallest to largest).

• The plot should look like a straight line with positive slope. It should not be S-shaped. It can be a bit
wobbly in the extreme tails, but that’s it.

• If the points on a normal quantile plot lie close to a straight line, the plot indicates that the data
appear to have come from a normal distribution.

• Outliers appear as points that are far away from the overall pattern of the plot.

74
Could Our Sample be From a Normal Distribution?

QQ Plot Prototype for the

Normal Distribution

75
 Assessing Normality
• If plots are OK → data could come from a normal distribution …

but it could come from some other distribution.

So good plots don’t prove data came from a normal distribution.

• If plots are not OK → data probably does not come from a normal distribution,

we can’t assume data is from normal population

76
 HistData-package {HistData}

Data sets from the History of Statistics and Data Visualization

77
Mean and SD of Population of Heights

78
Plot the Distribution of Heights

79
80
Normal Quantile Plot and Boxplot of Calories in Tim Horton’s Donuts
Right-skewed Distribution

81
Normal Quantile Plot and Boxplot of Educational Attainment Percentages
Left-skewed Distribution

82