4.1 4.2 4.3 4.

4.3 Correlation Coefficients

and Outliers
Roller coasters are popular rides at amusement parks. A recent survey
counted 1,797 roller coaster rides in the world. 734 of them are in North
America. Roller coasters differ in maximum drop, maximum height, track
length, ride time, and coaster type (wood or steel).

• Which roller coaster variables do you think are strongly related to the
top speed on the ride?

Problem 4.3
Statisticians measure the strength of a linear relationship between two variables
using a number called the correlation coefficient. This number is a decimal
between -1 and 1. When the points lie close to a straight line, the correlation
coefficient is close to -1 or 1.
continued on the next page >

Investigation 4 Variability and Associations in Numerical Data 87

 4.1 4.2
4.3 4.3
4.4 4.4
4.5

Problem 4.3 continued

• When points cluster close to a line with positive slope, the

correlation coefficient is almost 1, and with negative slope, the
correlation coefficient is almost -1.
• Points that do not cluster close to any line have a correlation
coefficient of almost 0.
• Positive association has correlation coefficients greater than 0
while negative association has correlation coefficients less than 0.

A 1. The graph below has a correlation coefficient of 1.0. What do you think
a correlation coefficient of 1.0 means?
A cc of 1.0 means the graph
has positive association that is
perfectly predictable.

2. Which of the six scatter plots below (a)–(f ) has a correlation coefficient
of -1.0? What do you think a correlation coefficient of -1 .0 means?
Letter b. It goes downward and it makes a straight line. (negative association)
a. 0.4 b. -1 c. 0.8

It has positive association and the It has positive association but its still
dots are more spread out. not 100% accurate. (still tight)
d. -0.4 e. -0.8 f. 0

It has negative association It has negative There is no

and the dots are more association but its not 100% pattern
spread out. accurate. (still tight) whatsoever.
3. Match correlation coefficients -0.8, -0.4, 0.0, 0.4, and 0.8 with the
other five scatter plots. Explain your reasoning.
continued on the next page >

88 Thinking With Mathematical Models

 4.1 4.2
4.3 4.3
4.4 4.4
4.5

Problem 4.3 continued

When you inspect a scatter plot, often you are looking for a strong association
between the variables.

B The scatter plot below shows the relationship between the top speed of a
roller coaster and its maximum drop. The pink dots represent wood-frame
roller coasters. The blue dots represent steel-frame coasters.

120
110
100
90
80
Top Speed (mph)

70
60
50
40
30
20
10
0
0 40 80 120 160 200 240 280 320 360 400 440
Maximum Drop (ft)

Yes, there is a strong 1. Suppose you drew one linear model for all the data in the graph.
positive association
between max drop and Could you use the model to make an accurate prediction about the
max speed. top speed of the roller coaster with a given maximum drop? Explain.
2. Estimate the correlation coefficient for the top speed and the
maximum drop. Is the correlation coefficient closest to -1, -0.5, 0,
0.5, or 1? Closest to 1
3. Is the maximum drop of a roller coaster likely to be one of the causes
of the top speed of the coaster? Why or why not?
Yes because the higher the coaster, the faster it will go going down.
continued on the next page >

Investigation 4 Variability and Associations in Numerical Data 89

 4.1 4.2
4.3 4.3
4.4 4.4
4.5

Problem 4.3 continued

C The scatter plot below shows the relationship between the top speed of
a roller coaster and its track length. The pink dots represent wood-frame
roller coasters. The blue dots represent steel-frame coasters.

120
110
100
90
80
Top Speed (mph)

70
60
50
40
30
20
10
0
0 1000 2000 3000 4000 5000 6000 7000
Track Length (ft)

1. Suppose you drew one linear model for all the data in the graph. Yes, however I cannot
Could you use the model to make an accurate prediction about the predict with 100%
certainty.
top speed of the roller coaster with a given track length? Explain.
2. Estimate the correlation coefficient for the top speed and track length.
Is the correlation coefficient closest to -1, -0.5, 0, 0.5, or 1? Closest to 0.5
3. Is the track length of a roller coaster likely to be one of the causes of
the top speed of the coaster? Why or why not? No, because the points of the graph are
scattered. It doesn't look like the length of the
coaster affects the speed.
4. Computer and calculator data analysis tools can take data pairs like
those plotted above and calculate exact correlation coefficients. Use
the tool that you have available to find the correlation coefficient for
the sample of (track length, top speed) data in the table.

Track Length (ft) 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000
Top Speed (mph) 5 20 45 50 40 50 55 60 85 65

continued on the next page >

90 Thinking With Mathematical Models

 4.1 4.2
4.3 4.3
4.4 4.4
4.5

Problem 4.3 continued

D The scatter plot below shows the relationship between the top speed of a
roller coaster and the ride time. The pink dots represent wood-frame roller
coasters. The blue dots represent steel-frame coasters.

120
110
100
90
80
Top Speed (mph)

70
60
50
40
30
20
10
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Ride Time (min)

Yes, but it is 1. Suppose you drew one linear model for all the data in the graph.
difficult to accurately
predict exact Could you use the model to make an accurate prediction about the
coordinates. top speed of the roller coaster with a given ride time? Explain.
2. Estimate the correlation coefficient for the top speed and ride time. Is
the correlation coefficient closest to -1, - 0.5, 0, 0.5, or 1? 0.5
3. Suppose most of the points on a scatter plot cluster near a line, with
only a few that don’t fit the pattern. The points that lie outside a cluster
are called outliers. Use the graph above. Find each point. Then decide
whether the point is an outlier. If it is, explain why you think it is an
outlier. Far from cluster
a. (1.75, 50) b. (0.30, 80) c. (3.35, 75)
d. (0.28, 120) e. (0.80, 21) f. (1.0, 10) Not close to the
Farthest from cluster model line
g. Use the scatter plot in Question C. Find two outliers on that graph
and estimate their coordinates (track length, top speed).
(4,80) (3.7, 65) continued on the next page >

Investigation 4 Variability and Associations in Numerical Data 91

 4.1 4.2
4.3 4.3
4.4 4.4
4.5

Problem 4.3 continued

E The scatter plot shows the number of roller coaster riders and their ages on
a given day. The pink dots represent wood-frame roller coasters. The blue
dots represent steel-frame coasters.
On that day, forty-four 15-year-olds rode one of the roller coasters. The data
point is (15, 44).

90
80
70
Number of Riders

60
50
40
30
20
10
0
10 20 30 40 50 60 70 80
Age of Riders

Yes, there is definitely a

strong correlation with 1. Suppose you drew one linear model for all the data in the graph.
only few outliers. Could you use the model to make an accurate prediction about the
No, although the association number of riders on the roller coaster with a given age? Explain.
is correlative, age is not
causal. 2. Is the age of riders on a roller coaster likely to be one of the causes of
the number of riders on the coaster? Why or why not?

It is closest to -1. 3. Estimate the correlation coefficient for the number of riders and age of
riders. Is the correlation coefficient closest to -1, -0.5, 0, 0.5, or 1?
4. Are any of the data points outliers? If so, estimate the coordinates of
those points. (80,1)
(77,3)
(15,44)
F Is it possible to have a correlation coefficient close to - 1 or 1 with only a
few outliers? Explain your thinking. Yes, because a few outliers do not affect the
overall association.

Homework starts on page 96.

92 Thinking With Mathematical Models