Looking at Data—Distributions

IPS Chapter 1

 1.1: Displaying distributions with graphs

 1.2: Describing distributions with numbers

 1.3: Density Curves and Normal Distributions

© 2012 W.H. Freeman and Company

Looking at Data—Distributions
1.1 Displaying Distributions
with Graphs

© 2012 W.H. Freeman and Company

Objectives
1.1 Displaying distributions with graphs

 Variables
 Types of variables
 Graphs for categorical variables
 Bar graphs
 Pie charts
 Graphs for quantitative variables
 Histograms
 Stemplots
 Stemplots versus histograms
 Interpreting histograms
 Time plots
 Variables
In a study, we collect information—data—from cases. Cases can be
individuals, companies, animals, plants, or any object of interest.

A label is a special variable used in some data sets to distinguish the

different cases.

A variable is any characteristic of an case. A variable varies among

cases.

Example: age, height, blood pressure, ethnicity, leaf length, first language

The distribution of a variable tells us what values the variable takes and
how often it takes these values.
Two types of variables
 Variables can be either quantitative…

 Something that takes numerical values for which arithmetic operations,

such as adding and averaging, make sense.
 Example: How tall you are, your age, your blood cholesterol level, the
number of credit cards you own.

 … or categorical.

 Something that falls into one of several categories. What can be counted
is the count or proportion of cases in each category.
 Example: Your blood type (A, B, AB, O), your hair color, your ethnicity,
whether you paid income tax last tax year or not.
How do you know if a variable is categorical or quantitative?
Ask:
 What are the n cases/units in the sample (of size “n”)?

 What is being recorded about those n cases/units?

 Is that a number ( quantitative) or a statement ( categorical)?

Categorical Quantitative
Each individual is Each individual is
assigned to one of attributed a
Label several categories. numerical value.

Individuals DIAGNOSIS AGE AT DEATH

in sample
Patient A Heart disease 56
Patient B Stroke 70
Patient C Stroke 75
Patient D Lung cancer 60
Patient E Heart disease 80
Patient F Accident 73
Patient G Diabetes 69
Ways to chart categorical data
Because the variable is categorical, the data in the graph can be
ordered any way we want (alphabetical, by increasing value, by year,
by personal preference, etc.)

 Bar graphs
Each category is
represented by
a bar.

 Pie charts
The slices must
represent the parts of one whole.
Example: Top 10 causes of death in the United States 2006
% of top
Rank Causes of death Counts % of total deaths
10s
1 Heart disease 631,636 34% 26%
2 Cancer 559,888 30% 23%
3 Cerebrovascular 137,119 7% 6%
4 Chronic respiratory 124,583 7% 5%
5 Accidents 121,599 7% 5%
6 Diabetes mellitus 72,449 4% 3%
7 Alzheimer’s disease 72,432 4% 3%
8 Flu and pneumonia 56,326 3% 2%
9 Kidney disorders 45,344 2% 2%
10 Septicemia 34,234 2% 1%

All other causes 570,654 24%

For each individual who died in the United States in 2006, we record what was
the cause of death. The table above is a summary of that information.
Bar graphs
Each category is represented by one bar. The bar’s height shows the count (or
sometimes the percentage) for that particular category.

700,000

Top 10 causes of deaths in the United States 2006

600,000

500,000

400,000
The number of individuals
300,000 who died of an accident in
2006 is approximately
200,000 121,000.

100,000

0
 Top 10 causes of deaths in the United States 2006
700,000

600,000 Bar graph sorted by rank This Bar

500,000
 Easy to analyze graph is an
400,000 example of a
300,000 Pareto Chart
200,000

100,000

Sorted alphabetically
 Much less useful
Pie charts
Each slice represents a piece of one whole. The size of a slice depends on what
percent of the whole this category represents.
Percent of people dying from
top 10 causes of death in the United States in 2006
 Make sure your
labels match
the data.

Make sure
all percents
add up to 100.
Percent of deaths from top 10 causes

Percent of
deaths from
all causes
Child poverty before and after
government intervention—UNICEF, 2005

What does this chart tell you?

•The United States and Mexico have the highest
rate of child poverty among OECD (Organization
for Economic Cooperation and Development)
nations (22% and 28% of under 18).

•Their governments do the least—through taxes

and subsidies—to remedy the problem (size of
orange bars and percent difference between
orange/blue bars).

Identify the Pareto Chart in this graph.

Could you transform this bar graph to fit in 1

pie chart? In two pie charts? Why?
The poverty line is defined as 50% of national median
income.
Ways to chart quantitative data
 Stemplots
Also called a stem-and-leaf plot. Each observation is represented by a stem,
consisting of all digits except the final one, which is the leaf.

 Histograms
A histogram breaks the range of values of a variable into classes and
displays only the count or percent of the observations that fall into each
class.

 Line graphs: time plots

A time plot of a variable plots each observation against the time at which it
was measured.
Stem plots

How to make a stemplot: STEM LEAVES

 Separate each observation into a stem, consisting of

all but the final (rightmost) digit, and a leaf, which is
that remaining final digit. Stems may have as many
digits as needed, but each leaf contains only a single
digit.

 Write the stems in a vertical column with the smallest

value at the top, and draw a vertical line at the right
of this column.

 Write each leaf in the row to the right of its stem, in

increasing order out from the stem.
State Percent State Percent
Alabama
Alaska
1.5
4.1
Maine
W estVirginia
0.7
0.7 Percent of Hispanic residents
Arizona 25.3 Vermont 0.9
Arkansas
California
2.8
32.4
NorthDakota
Mississippi
1.2
1.3
in each of the 50 states
Colorado 17.1 SouthDakota 1.4
Connecticut 9.4 Alabama 1.5
Delaware 4.8 Kentucky 1.5
Florida 16.8 NewHampshire 1.7
Georgia 5.3 Ohio 1.9
Hawaii 7.2 Montana 2
Idaho 7.9 Tennessee 2
Illinois 10.7 Missouri 2.1
Indiana 3.5 Louisiana 2.4
Iowa 2.8 SouthCarolina 2.4
Kansas 7 Arkansas 2.8
Kentucky 1.5 Iowa 2.8
Louisiana 2.4 Minnesota 2.9
Maine 0.7 Pennsylvania 3.2
Maryland 4.3 Michigan 3.3
Massachusetts 6.8 Indiana 3.5
Michigan 3.3 W isconsin 3.6
Minnesota 2.9 Alaska 4.1
Mississippi 1.3
Step 1: Maryland 4.3 Step 2:
Missouri 2.1 NorthCarolina 4.7
Montana 2 Virginia 4.7
Nebraska 5.5 Delaware 4.8
Nevada 19.7 Oklahoma 5.2
NewHampshire 1.7 Georgia 5.3
NewJersey 13.3
Sort the Nebraska 5.5 Assign the
NewMexico 42.1 W yoming 6.4
NewYork
NorthCarolina
15.1
4.7 data Massachusetts 6.8 values to
Kansas 7
NorthDakota
Ohio
1.2
1.9
Hawaii
W ashington
7.2
7.2
stems and
Oklahoma
Oregon
5.2
8
Idaho
Oregon
7.9
8
leaves
Pennsylvania 3.2 RhodeIsland 8.7
RhodeIsland 8.7 Utah 9
SouthCarolina 2.4 Connecticut 9.4
SouthDakota 1.4 Illinois 10.7
Tennessee 2 NewJersey 13.3
Texas 32 NewYork 15.1
Utah 9 Florida 16.8
Vermont 0.9 Colorado 17.1
Virginia 4.7 Nevada 19.7
W ashington 7.2 Arizona 25.3
W estVirginia 0.7 Texas 32
W isconsin 3.6 California 32.4
W yoming 6.4 NewMexico 42.1
Stem Plot
 To compare two related distributions, a back-to-back stem plot with
common stems is useful.

 Stem plots do not work well for large datasets.

 When the observed values have too many digits, trim the numbers
before making a stem plot.

 When plotting a moderate number of observations, you can split

each stem.
Histograms
The range of values that a
variable can take is divided
into equal size intervals.

The histogram shows the

number of individual data
points that fall in each
interval.

The first column represents all states with a Hispanic percent in their
population between 0% and 4.99%. The height of the column shows how
many states (27) have a percent in this range.

The last column represents all states with a Hispanic percent in their
population between 40% and 44.99%. There is only one such state: New
Mexico, at 42.1% Hispanics.
Stemplots versus histograms

Stemplots are quick and dirty histograms that can easily be done by
hand, and therefore are very convenient for back of the envelope
calculations. However, they are rarely found in scientific or laymen
publications.
Interpreting histograms
When describing the distribution of a quantitative variable, we look for the
overall pattern and for striking deviations from that pattern. We can describe
the overall pattern of a histogram by its shape, center, and spread.

Histogram with a line connecting Histogram with a smoothed curve

each column  too detailed highlighting the overall pattern of
the distribution
Most common distribution shapes Symmetric
distribution

 A distribution is symmetric if the right and left

sides of the histogram are approximately mirror
images of each other.

 A distribution is skewed to the right if the right

side of the histogram (side with larger values)
extends much farther out than the left side. It is Skewed
skewed to the left if the left side of the histogram distribution

extends much farther out than the right side.

Complex,
multimodal
distribution

 Not all distributions have a simple overall shape,

especially when there are few observations.
 Outliers
An important kind of deviation is an outlier. Outliers are observations
that lie outside the overall pattern of a distribution. Always look for
outliers and try to explain them.

The overall pattern is fairly

symmetrical except for 2
states that clearly do not
belong to the main trend.
Alaska and Florida have
unusual representation of
the elderly in their
Alaska Florida
population.

A large gap in the

distribution is typically a
sign of an outlier.
How to create a histogram

It is an iterative process – try and try again.

What bin size should you use?

 Not too many bins with either 0 or 1 counts

 Not overly summarized that you lose all the information
 Not so detailed that it is no longer summary

 rule of thumb: start with 5 to 10 bins

Look at the distribution and refine your bins
(There isn’t a unique or “perfect” solution)
 Same data set

Not
summarized
enough

Too summarized
 Histogram of dry days in 1995

IMPORTANT NOTE:
Your data are the way they are.
Do not try to force them into a
particular shape.

It is a common misconception
that if you have a large enough
data set, the data will eventually
turn out nice and symmetrical.
Line graphs: time plots
In a time plot, time always goes on the horizontal, x axis.

We describe time series by looking for an overall pattern and for striking
deviations from that pattern. In a time series:

A trend is a rise or fall that

persists over time, despite
small irregularities.

A pattern that repeats itself

at regular intervals of time is
called seasonal variation.
Retail price of fresh oranges
over time

Time is on the horizontal, x axis.

The variable of interest—here

“retail price of fresh oranges”—
goes on the vertical, y axis.

This time plot shows a regular pattern of yearly variations. These are seasonal
variations in fresh orange pricing most likely due to similar seasonal variations in
the production of fresh oranges.

There is also an overall upward trend in pricing over time. It could simply be
reflecting inflation trends or a more fundamental change in this industry.
 A time plot can be used to compare two or more
data sets covering the same time period.

1918 influenza epidemic

1918 influenza epidemic
Date # Cases # Deaths
week 1 36 0 1918 influenza epidemic
week 2 531 0 10000 800
week 3 4233 130 9000
10000 800 700

# cases diagnosed
week 4 8682 552 8000

# deaths reported
9000 700 600
week 5 7164 738 7000
8000
week 6 2229 414 6000 600 500
Incidence 7000
week 7 600 198 5000
6000 500 400
week 8 164 90 4000
5000 400 300
4000
3000 300 200
week 9 57 56 3000
2000 200 100
week 10 722 50 2000
1000
week 11 1517 71 1000 0 100
0
week 12 1828 137 0 0
week 13 1539 178
1

11

13

15

17
1 3 5 7 9 11 13 15 17
k

k
ee

ee

ee

ee

ee

k
week 14 2416 194 k k k k k

ee

ee

ee

ee
ee ee e e e k k k k
w

w
e e e ee ee ee ee

w
week 15 3148 290 w w w w w w w w w
week 16 3465 310
# Cases # Deaths
week 17 1440 149 # Cases # Deaths

The pattern over time for the number of flu diagnoses closely resembles that for the
number of deaths from the flu, indicating that about 8% to 10% of the people
diagnosed that year died shortly afterward, from complications of the flu.
Scales matter Death rates from cancer (US, 1945-95)

How you stretch the axes and choose your 250

scales can give a different impression.

Death rate (per thousand)

200
Death rates from cancer (US, 1945-95)
150
250
Death rate (per

200 100
thousand)

150
100 50
50
0 0
1940 1950 1960 1970 1980 1990 2000 1940 1960 1980 2000

Years Years

Death rates from cancer (US, 1945-95)

250 Death rates from cancer (US, 1945-95)

A picture is worth a
200
220
thousand words,
Death rate (per thousand)
Death rate (per thousand)

200
150
180
BUT
100 160

140 There is nothing like

50

120
hard numbers.
0
1940 1960 1980 2000  Look at the scales.
1940 1960 1980 2000 Years
Years
Why does it matter?

What's wrong with

these graphs?
Cornell’s tuition over time
Careful reading
reveals that:

1. The ranking graph covers an 11-year period, the

tuition graph 35 years, yet they are shown
comparatively on the cover and without a
horizontal time scale.
2. Ranking and tuition have very different units, yet
both graphs are placed on the same page without
a vertical axis to show the units.
Cornell’s ranking over time
3. The impression of a recent sharp “drop” in the
ranking graph actually shows that Cornell’s rank
has IMPROVED from 15th to 6th ...
Looking at Data—Distributions
1.2 Describing distributions
with numbers

© 2012 W.H. Freeman and Company

Objectives
1.2 Describing distributions with numbers

 Measures of center: mean, median

 Mean versus median

 Measures of spread: quartiles, standard deviation

 Five-number summary and boxplot

 Choosing among summary statistics

 Changing the unit of measurement

Measure of center: the mean
The mean or arithmetic average

To calculate the average, or mean, add 58.2 64.0

all values, then divide by the number of 59.5 64.5
60.7 64.1
cases. It is the “center of mass.”
60.9 64.8
61.9 65.2
61.9 65.7
62.2 66.2
Sum of heights is 1598.3
62.2 66.7
divided by 25 women = 63.9 inches 62.4 67.1
62.9 67.8
63.9 68.9
63.1 69.6
63.9
woman height woman height Mathematical notation:
(i) (x) (i) (x)
i=1 x1= 58.2 i = 14 x14= 64.0
x1  x 2  ... x n
x
i=2 x2= 59.5 i = 15 x15= 64.5
i=3 x3= 60.7 i = 16 x16= 64.1
n
i=4 x4= 60.9 i = 17 x17= 64.8
i=5 x5= 61.9 i = 18 x18= 65.2 n
1
i=6 x6= 61.9 i = 19 x19= 65.7 x   xi
i=7 x7= 62.2 i = 20 x20= 66.2 n i1
i=8 x8= 62.2
x
i = 21 = 66.7
21

i=9 x9= 62.4 i = 22 x22= 67.1

i = 10 x10= 62.9 i = 23 x23= 67.8
1598.3
i = 11 x11= 63.9 i = 24 x24= 68.9 x  63.9
i = 12 x12= 63.1 x
i = 25 = 69.6
25
25
i = 13 x13= 63.9 n=25 S=1598.3

Learn right away how to get the mean using your calculators.
 Your numerical summary must be meaningful.
Height of 25 women in a class

x  63.9 The distribution of women’s

heights appears coherent and
symmetrical. The mean is a good
numerical summary.


Here the shape of

x  69.6
the distribution is
wildly irregular.
Why?

Could we have 
more than one
plant species or
phenotype?
 Height of Plants by Color

5
x  63.9 x  70.5 x  78.3
red
4 pink
Number of Plants

   blue

0
58 60 62 64 66 68 70 72 74 76 78 80 82 84
Height in centimeters

A single numerical summary here would not make sense.

Measure of center: the median
The median is the midpoint of a distribution—the number such
that half of the observations are smaller and half are larger.
1 1 0.6 1 1 0.6
2 2 1.2
1. Sort observations by size. 2 2 1.2
3 3 1.6 n = number of observations 3 3 1.6
4 4 1.9 4 4 1.9
______________________________
5 5 1.5 5 5 1.5
6 6 2.1 6 6 2.1
7 7 2.3 7 7 2.3
2.a. If n is odd, the median is
8 8 2.3 8 8 2.3
9 9 2.5 observation (n+1)/2 down the list 9 9 2.5
10 10 2.8 10 10 2.8
11 11 2.9  n = 25 11 11 2.9
12 12 3.3 (n+1)/2 = 26/2 = 13 12 3.3
13 3.4 13 3.4
14 1 3.6 Median = 3.4 14 1 3.6
15 2 3.7 15 2 3.7
16 3 3.8 16 3 3.8
17 4 3.9 2.b. If n is even, the median is the 17 4 3.9
18 5 4.1 mean of the two middle observations. 18 5 4.1
19 6 4.2 19 6 4.2
20 7 4.5
n = 24  20 7 4.5
21 8 4.7 21 8 4.7
22 9 4.9 n/2 = 12 22 9 4.9
23 10 5.3 Median = (3.3+3.4) /2 = 3.35 23 10 5.3
24 11 5.6 24 11 5.6
25 12 6.1
Comparing the mean and the median
The mean and the median are the same only if the distribution is
symmetrical. The median is a measure of center that is resistant to skew
and outliers. The mean is not.

Mean and median for a

symmetric distribution

Mean
Median

Mean and median for

skewed distributions

Left skew Mean Mean Right skew

Median Median
Mean and median of a distribution with outliers

x  3.4 x  4.2
Percent of people dying
Without the outliers
  With the outliers

The mean is pulled to the The median, on the other hand,

right a lot by the outliers is only slightly pulled to the right
(from 3.4 to 4.2). by the outliers (from 3.4 to 3.6).
 Impact of skewed data

Symmetric distribution…

x  3.4
Disease X:
M  3.4

Mean and median are the same.



… and a right-skewed distribution

Multiple myeloma:
x  3.4
M  2.5

The mean is pulled toward

the skew.

Measure of spread: the quartiles
1 1 0.6
2 2 1.2
3 3 1.6
The first quartile, Q1, is the value in the 4 4 1.9
5 5 1.5
sample that has 25% of the data at or 6 6 2.1
7 7 2.3 Q1= first quartile = 2.2
below it ( it is the median of the lower 8 1 2.3
9 2 2.5
half of the sorted data, excluding M). 10 3 2.8
11 4 2.9
12 5 3.3
M = median = 3.4 13 3.4
14 1 3.6
15 2 3.7
16 3 3.8
17 4 3.9
The third quartile, Q3, is the value in the 18 5 4.1
19 6 4.2
sample that has 75% of the data at or 20 7 4.5
Q3= third quartile = 4.35
21 1 4.7
below it ( it is the median of the upper 22 2 4.9
23 3 5.3
half of the sorted data, excluding M). 24 4 5.6
25 5 6.1
Five-number summary and boxplot
25 6 6.1 Largest = max = 6.1
24 5 5.6
23 4 5.3 BOXPLOT
22 3 4.9
7
21 2 4.7
20 1 4.5 Q3= third quartile
19 6 4.2
6
18 5 4.1
= 4.35
17 4 3.9 5

Years until death

16 3 3.8
15 2 3.7 4
14 1 3.6
13 3.4 3
12 6 3.3
M = median = 3.4
11 5 2.9 2
10 4 2.8
9 3 2.5 1
8 2 2.3
7 1 2.3 0
6 6 2.1 Q1= first quartile Disease X
5 5 1.5 = 2.2
4 4 1.9
3 3 1.6 Five-number summary:
2 2 1.2
min Q1 M Q3 max
1 1 0.6 Smallest = min = 0.6
 Boxplots for skewed data

Comparing box plots for a normal

and a right-skewed distribution
15
14
13
12
Years until death

11
10 Boxplots remain
9
8 true to the data and
7
6 depict clearly
5
4 symmetry or skew.
3
2
1
0
Disease X Multiple Myeloma
Suspected outliers

Outliers are troublesome data points, and it is important to be able to

identify them.

One way to raise the flag for a suspected outlier is to compare the
distance from the suspicious data point to the nearest quartile (Q1 or
Q3). We then compare this distance to the interquartile range
(distance between Q1 and Q3).

We call an observation a suspected outlier if it falls more than 1.5

times the size of the interquartile range (IQR) below the first quartile or
above the third quartile. This is called the “1.5 * IQR rule for outliers.”
25 6 7.9
24 5 6.1
23 4 5.3 8
22 3 4.9
21 2 4.7 7
20 1 4.5
Distance to Q3
19 6 4.2 Q3 = 4.35 6 7.9 − 4.35 = 3.55
18 5 4.1
17 4 3.9 5

Years until death

16 3 3.8
15 2 3.7 4 Interquartile range
14 1 3.6 Q3 – Q1
3
13 3.4 4.35 − 2.2 = 2.15
12 6 3.3
11 5 2.9 2
10 4 2.8
9 3 2.5 1
8 2 2.3
7 1 2.3 0
Q1 = 2.2
6 6 2.1 Disease X
5 5 1.5
4 4 1.9
3 3 1.6 Individual #25 has a value of 7.9 years, which is 3.55 years above
2 2 1.2
1 1 0.6
the third quartile. This is more than 3.225 years, 1.5 * IQR. Thus,
individual #25 is a suspected outlier.
Measure of spread: the standard deviation
The standard deviation “s” is used to describe the variation around the
mean. Like the mean, it is not resistant to skew or outliers.

1. First calculate the variance s2.

n
1
s 
2
 (x
n 1 1 i
 x ) 2

2. Then take the square root to get

x
 the standard deviation s.

1 n
Mean
s 
n 1 1
( xi  x ) 2

±1
s.d.
 Calculations … Women’s height (inches)

n
1
s
df
 (x
1
i  x) 2

Mean = 63.4
Sum of squared deviations from mean = 85.2
Degrees freedom (df) = (n − 1) = 13
s2 = variance = 85.2/13 = 6.55 inches squared
s = standard deviation = √6.55 = 2.56 inches

We’ll never calculate these by hand, so make sure to know how to

get the standard deviation using your calculator or software.
Variance and Standard Deviation
 Why do we square the deviations?
 The sum of the squared deviations of any set of observations from their
mean is the smallest that the sum of squared deviations from any
number can possibly be.
 The sum of the deviations of any set of observations from their mean is
always zero.

 Why do we emphasize the standard deviation rather than the

variance?
 s, not s2, is the natural measure of spread for Normal distributions.
 s has the same unit of measurement as the original observations.

 Why do we average by dividing by n − 1 rather than n in calculating

the variance?
 The sum of the deviations is always zero, so only n − 1 of the squared
deviations can vary freely.
 The number n − 1 is called the degrees of freedom.
Properties of Standard Deviation
 s measures spread about the mean and should be used only when
the mean is the measure of center.

 s = 0 only when all observations have the same value and there is
no spread. Otherwise, s > 0.

 s is not resistant to outliers.

 s has the same units of measurement as the original observations.

Software output for summary statistics:

Excel - From Menu:

Tools/Data Analysis/
Descriptive Statistics

Give common
statistics of your
sample data.

Minitab
Choosing among summary statistics

 Because the mean is not Height of 30 Women

resistant to outliers or skew, use
69
it to describe distributions that are 68
fairly symmetrical and don’t have 67
66
outliers.

Height in Inches
65
 Plot the mean and use the 64

standard deviation for error bars. 63

62
61
60
 Otherwise use the median in the
59
five number summary which can 58

be plotted as a boxplot. Box Plot

Boxplot Mean ±
Mean +/- SD
SD
What should you use, when, and why?

Arithmetic mean or median?

 Middletown is considering imposing an income tax on citizens. City hall

wants a numerical summary of its citizens’ income to estimate the total tax
base.
 Mean: Although income is likely to be right-skewed, the city government
wants to know about the total tax base.

 In a study of standard of living of typical families in Middletown, a sociologist

makes a numerical summary of family income in that city.

 Median: The sociologist is interested in a “typical” family and wants to

lessen the impact of extreme incomes.
Changing the unit of measurement
Variables can be recorded in different units of measurement. Most
often, one measurement unit is a linear transformation of another
measurement unit: xnew = a + bx.

Temperatures can be expressed in degrees Fahrenheit or degrees Celsius.

TemperatureFahrenheit = 32 + (9/5)* TemperatureCelsius  a + bx.

Linear transformations do not change the basic shape of a distribution

(skew, symmetry, multimodal). But they do change the measures of
center and spread:
Multiplying each observation by a positive number b multiplies both
measures of center (mean, median) and spread (IQR, s) by b.
Adding the same number a (positive or negative) to each observation
adds a to measures of center and to quartiles but it does not change
measures of spread (IQR, s).
Looking at Data—Distributions
1.3 Density Curves and
Normal Distributions

© 2012 W.H. Freeman and Company

Objectives
1.3 Density curves and Normal distributions

 Density curves

 Measuring center and spread for density curves

 Normal distributions

 The 68-95-99.7 rule

 Standardizing observations

 Using the standard Normal Table

 Inverse Normal calculations

 Normal quantile plots

Density curves
A density curve is a mathematical model of a distribution.
The total area under the curve, by definition, is equal to 1, or 100%.
The area under the curve for a range of values is the proportion of all
observations for that range.

Histogram of a sample with the

smoothed, density curve
describing theoretically the
population.
Density curves come in any
imaginable shape.

Some are well known

mathematically and others aren’t.
Median and mean of a density curve
The median of a density curve is the equal-areas point: the point that
divides the area under the curve in half.

The mean of a density curve is the balance point, at which the curve
would balance if it were made of solid material.

The median and mean are the same for a symmetric density curve.
The mean of a skewed curve is pulled in the direction of the long tail.
Normal distributions
Normal – or Gaussian – distributions are a family of symmetrical, bell-
shaped density curves defined by a mean m (mu) and a standard
deviation s (sigma) : N(m,s).

2
1 x  m 
1  
2  s 

f (x)  e
s 2

 x x

e = 2.71828… The base of the natural logarithm

π = pi = 3.14159…
 A family of density curves

Here, means are the same (m = 15)

while standard deviations are
different (s = 2, 4, and 6).

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30

Here, means are different

(m = 10, 15, and 20) while standard
deviations are the
same (s = 3).

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
The 68-95-99.7% Rule for Normal Distributions

 About 68% of all observations

are within 1 standard deviation Inflection point

(s)of the mean (m).

 About 95% of all observations

are within 2 s of the mean m.

 Almost all (99.7%) observations

are within 3 s of the mean.
mean µ = 64.5 standard deviation s = 2.5
N(µ, s) = N(64.5, 2.5)

Reminder: µ (mu) is the mean of the idealized curve, while x is the mean of a sample.
σ (sigma) is the standard deviation of the idealized curve, while s is the s.d. of a sample.
The standard Normal distribution
Because all Normal distributions share the same properties, we can
standardize our data to transform any Normal curve N(m,s) into the
standard Normal curve N(0,1).

N(64.5, 2.5) N(0,1)

=>

x z
Standardized height (no units)

For each x we calculate a new value, z (called a z-score).

Standardizing: calculating z-scores
A z-score measures the number of standard deviations that a data
value x is from the mean m.
When x is 1 standard deviation larger
(x  m ) than the mean, then z = 1.
z m s  m s
s for x  m  s , z 
s
 1
s

When x is 2 standard deviations larger

than the mean, then z = 2.
m  2s  m 2s
for x  m  2s , z   2
s s

When x is larger than the mean, z is positive.

When x is smaller than the mean, z is negative.
Ex. Women heights N(µ, s) =
N(64.5, 2.5)
Women’s heights follow the N(64.5”,2.5”)
distribution. What percent of women are Area= ???

shorter than 67 inches tall (that’s 5’6”)? Area = ???

mean µ = 64.5"
standard deviation s = 2.5"
m = 64.5” x = 67”
x (height) = 67" z=0 z=1

We calculate z, the standardized value of x:

(x  m) (67  64.5) 2.5

z , z   1  1 stand. dev. from mean
s 2.5 2.5

Because of the 68-95-99.7 rule, we can conclude that the percent of women
shorter than 67” should be, approximately, .68 + half of (1 - .68) = .84 or 84%.
 Using the standard Normal table
Table A gives the area under the standard Normal curve to the left of any z value.

.0082 is the
area under
N(0,1) left
of z = -
2.40

0.0069 is the area

.0080 is the area
under N(0,1) left
under N(0,1) left
(…) of z = -2.46
of z = -2.41
Percent of women shorter than 67”

For z = 1.00, the area under

the standard Normal curve
to the left of z is 0.8413.

N(µ, s) =
N(64.5”, 2.5”)

Area ≈ 0.84

Conclusion: Area ≈ 0.16

84.13% of women are shorter than 67”.
By subtraction, 1 - 0.8413, or 15.87% of
women are taller than 67". m = 64.5” x = 67”
z=1
Tips on using Table A
Because the Normal distribution is symmetrical, there are 2 ways that
you can calculate the area under the standard Normal curve to the
right of a z value.

Area = 0.9901

Area = 0.0099

z = -2.33

area right of z = area left of -z

area right of z = 1 - area left of z

Tips on using Table A
To calculate the area between 2 z- values, first get the area under N(0,1)
to the left for each z-value from Table A.

Then subtract the

smaller area from the
larger area.

A common mistake made by

students is to subtract both z
values. But the Normal curve
is not uniform. area between z1 and z2 =
area left of z1 – area left of z2

 The area under N(0,1) for a single value of z is zero.

(Try calculating the area to the left of z minus that same area!)
 The National Collegiate Athletic Association (NCAA) requires Division I athletes to
score at least 820 on the combined math and verbal SAT exam to compete in their
first college year. The SAT scores of 2003 were approximately normal with mean
1026 and standard deviation 209.

What proportion of all students would be NCAA qualifiers (SAT ≥ 820)?

x  820
m  1026
s  209
(x  m)
z
s
(820 1026)
z
209 area right of 820 = total area - area left of 820
206 = 1 - 0.1611
z  0.99 ≈ 84%
209
Table A : area under
N(0,1) to the left of Note: The actual data may contain students who scored
z = -0.99 is 0.1611 exactly 820 on the SAT. However, the proportion of scores
or approx. 16%. exactly equal to 820 is 0 for a normal distribution is a
consequence of the idealized smoothing of density curves.
 The NCAA defines a “partial qualifier” eligible to practice and receive an athletic
scholarship, but not to compete, with a combined SAT score of at least 720.

What proportion of all students who take the SAT would be partial qualifiers?
That is, what proportion have scores between 720 and 820?

x  720
m  1026
s  209
(x  m)
z
s
(720 1026)
z
209 area between = area left of 820 - area left of 720
306 720 and 820 = 0.1611 - 0.0721
z  1.46
209 ≈ 9%
Table A : area under
N(0,1) to the left of
z = -1.46 is 0.0721
About 9% of all students who take the SAT have scores
or approx. 7%. between 720 and 820.
 The cool thing about working with
normally distributed data is that
we can manipulate it, and then
find answers to questions that
involve comparing seemingly non-
comparable distributions.

We do this by “standardizing” the

N(0,1)
data. All this involves is changing
the scale so that the mean now = 0
and the standard deviation =1. If
you do this to different distributions
it makes them comparable.
(x  m )
z
s
Ex. Gestation time in malnourished mothers
What is the effect of better maternal care on gestation time and preemies?

The goal is to obtain pregnancies 240 days (8 months) or longer.

What improvement did we get

by adding better food?

m266
s15

m250
s20

180 200 220 240 260 280 300 320

Gestation time (days)

Vitamins only Vitamins and better food

Under each treatment, what percent of mothers failed to carry their babies at
least 240 days?

Vitamins Only m=250, s=20,

x=240
x  240
m  250
s  20
(x  m)
z
s
(240  250)
z
20
10 170 190 210 230 250 270 290 310
z  0.5
20 Gestation time (days)
(half a standard deviation)
Table A : area under N(0,1) to Vitamins only: 30.85% of women
the left of z = -0.5 is 0.3085.
would be expected to have gestation
times shorter than 240 days.
 Vitamins and better food
m=266, s=15,
x=240
x  240
m  266
s  15
(x  m)
z
s
(240  266)
z
15
26 206 221 236 251 266 281 296 311
z  1.73
15 Gestation time (days)
(almost 2 sd from mean)
Table A : area under N(0,1) to Vitamins and better food: 4.18% of women

the left of z = -1.73 is 0.0418. would be expected to have gestation

times shorter than 240 days.

Compared to vitamin supplements alone, vitamins and better food resulted in a much
smaller percentage of women with pregnancy terms below 8 months (4% vs. 31%).
Inverse normal calculations
We may also want to find the observed range of values that correspond
to a given proportion/ area under the curve.

For that, we use Table A backward:

 we first find the desired

area/ proportion in the
body of the table,

 we then read the

corresponding z-value
from the left column and
top row.

For an area to the left of 1.25 % (0.0125),

the z-value is -2.24
Vitamins and better food
How long are the longest 75% of pregnancies when mothers with malnutrition are
given vitamins and better food?
m=266, s=15,
m  266 upper area 75%
s  15
upper area  75%
lower area  25%
upper 75%
x?
Table A : z value for the
lower area 25% under
N(0,1) is about - 0.67. ?
206 221 236 251 266 281 296 311
(x  m) Gestation time (days)
z  x  m  ( z *s )
s Remember that Table A gives the area to
x  266  (0.67 *15) the left of z. Thus, we need to search for
x  255.95  256 the lower 25% in Table A in order to get z.

 The 75% longest pregnancies in this group are about 256 days or longer.
Normal quantile plots
One way to assess if a distribution is indeed approximately normal is to
plot the data on a normal quantile plot.

The data points are ranked and the percentile ranks are converted to z-
scores with Table A. The z-scores are then used for the x axis against
which the data are plotted on the y axis of the normal quantile plot.

 If the distribution is indeed normal the plot will show a straight line,
indicating a good match between the data and a normal distribution.

 Systematic deviations from a straight line indicate a non-normal

distribution. Outliers appear as points that are far away from the overall
pattern of the plot.
 Good fit to a straight line: the Curved pattern: the data are not
distribution of rainwater pH normally distributed. Instead, it shows
values is close to normal. a right skew: a few individuals have
particularly long survival times.

Normal quantile plots are complex to do by hand, but they are standard
features in most statistical software.
 Alternate Slide

The following slide offers alternate software

output data and examples for this presentation.
Software output for summary statistics:

JMP- From Menu: 

Analyze/Distribution/
Heights  Y, Columns/OK

Give common statistics of

your sample data.

 CrunchIt!
Stat  Summary Statistics  Columns
Women’s Heights (inches)
n, Min, Q1, Median, Q3, Max, Mean, Std. Dev.
OK

(Control Click or Click Shift to select several statistics.)