2.2.
3 Visual Presentation: Stem-and-Leaf Plot, Dot Plot, and Scatter plot
In helping students to develop their data analysis skills, it is important to keep the following
three points in mind:
⇒ For most data sets, there is more than one appropriate and informative way to plot the
data.
⇒ It is valuable to see the same data plotted in several ways, both for the purpose of
learning more about the data, and because this will help students to gain experience in choosing
appropriate methods of data display and analysis. Professionals routinely do this in order to explore
patterns in their data.
⇒ The graphing styles introduced here and incorporated into the kit activities are only a
small fraction of those that could be used. Creative insights are an asset to good graphing and data
analysis. If a student invents a different kind of graph that successfully makes a point about the
data that can be interpreted by someone else, then the graph is useful.
[Link] Stem-and-Leaf Plot
A stem-and-leaf plot provides an alternative to a line plot or histogram for obtaining a picture of
the data distribution when data can be represented as integer (counting) numbers. The resulting
graph looks very much like a line plot or a histogram. However, this type of plot can be constructed
very quickly- it can be used, for example, when a teacher wants to construct a graph on a
chalkboard or an overhead using data collected by the children in the class. The children can simply
call out their numbers one by one and the teacher can enter the numbers onto the graph.
Example 5: Eric’s third period PE class just had a fitness exam. Each child did as many sit-ups as
he/she could do in one minute. Eric’s score was 45. How well did he do relative to the rest of the
class? The data for the class are as follows:
49 58 57 43 46 35 42 56 47 45 45 43 51 36 41 50 34 42 46 47 64 40 38 41 50 48
A stem-and-leaf plot provides a convenient way to display the distribution of these data. In
a stem-and-leaf graph, we separate the digits of each data point into the right most digit vs. the
rest, so, for example, the 1s place becomes the “leaf,” and the rest of the number, becomes the
“stem.” So in the above data set, we can split the numbers into the number of 10s (the stem) and
the number of 1s (the leaf). 49 thus becomes 40 + 9 or 4 tens plus 9 ones. We are not restricted to
2-digit numbers- for example, we could have collected forearm lengths in mm, in which case the
numbers above might be 100 greater, so 149 = 14 tens plus 9 ones, etc.
In constructing the stem-and-leaf plot we end up with something that looks very much like
a line plot or histogram. There are two differences, however. First, we pool counts within a class
of numbers, e.g. all counts 30-39, 40-49, etc. Second, we retain the actual numbers to make the
graph, rather than using dots or Xs, so that we can continue to use the graph for determining other
summaries of the data, such as the median. (Note that we lose this information when we use a
histogram, although we can still obtain such summaries from a line plot).
⇒ Note: to get a good picture of the data, the final plot must be made quite neatly. To encourage
this, it is often convenient to use graph paper, and to write each digit in a separate box on the graph
paper.
The method of construction:
Step #1 Find the largest and smallest scores (the largest is 64 and the smallest is 34). If one person
in the class is making the plot, this is easy to do orally by asking for a small/large number and then
asking for successively smaller/larger numbers until there are no more extreme values.
Step #2 Because the smallest number, 34, has a 3 in the tens place, and the largest number, 64, has
a 6 in the tens place, the stems will be the digits 3 to 6. For younger children you may want to
simply write down all the tens place digits (1-9). Write these digits vertically with a line to the
right.
Stem
3|
4|
5|
6|
Step #3 Separate each score into a stem (number of tens) and a leaf (the 1s digit). Write the leaf
on the plot next to the stem. The first score is 49. The 9 is placed next to the 4 in the stem:
Stem Leaf
3|
4| 9
5|
6|
Continue until all the scores (from left to right) in the above data set are listed on the plot:
3|5648
4|936275531267018
5|876100
6|4
Step #4 On a new plot, rearrange the leaves so that they are ordered from smallest to largest.
3|4568
4|011223355667789
5|001678
6|4
Step #5 Add a title and a key to the plot.
Sit-ups/min. in Period 3 PE class
3|4568
4|011223355667789
5|001678
6|4
3|4 represents 34 sit-ups
Interpreting the data:
The key descriptive terms, such as symmetry, skewness, mode, outliers, clusters and gaps,
can all be used with a stem-and-leaf plot, just as with a line plot or histogram. For this example,
scores are relatively symmetric with no unique mode. But since most of the scores are in the 40s,
we could call this the modal class or category. Where is Eric’s score? Right in the middle of this
modal category. Eric is “typical” of a student from this class.
Variations and extensions:
For children who have mastered the concept of place value it is possible to introduce
variants on the stem-and-leaf plot.
a) Split stems:
In the above example, there were many data points in the middle categories. Spreading the
data out might reveal additional patterns. One way to do this is to split the stems into a smaller
number of equally-sized units. For example, units of 5 ones (30-34, 35-39, 40- 44, etc.). Taking
the previous plot, we then get:
Sit-ups/min. in Period 3 PE class
3|4
• |568
4|0112233
• |55667789
5|001
• |678
6 |4
3|4 represents 34 sit-ups
• | 5 = 35 sit-ups
( • is same value as stem above it)
Can you see how this spread out the data? This question will be answered when we discuss the
different measures of variability.
b) Back-to-back stem-and-leaf plots:
Like line plots and histograms, stem-and-leaf plots can be used to compare two situations.
For example, there might be data for a second class, perhaps of older children, or from a class
which has been doing sit-ups every day for a couple of weeks:
Comparison of sit-ups/min in 2 classes
Class 2 Class 1
3 2 | 3 | 4
5 5| • | 5 6 8
3 2 0 | 4 | 0 1 1 2 2 3 3
887665| • | 5 5 6 6 7 7 8 9
5433200| 5 | 0 0 1
97655| • | 6 7 8
310| 6 | 4
| 3 | 4 represents 34 sit-ups
0 | 5 | represents 50 sit-ups
Which class is generally more fit?
⇒ Note that back-to-back stem-and-leaf plots require writing the left-hand numbers backwards,
and therefore this method of plotting is not suitable for the younger children.
⇒ Note also that an easy way to demonstrate the back-to-back version is to use 3 transparencies:
one containing only the stem, the second, overlaid on the first for the data from situation 1, and
the third, overlaid on the first (after removing the second) for the data on situation 2. Then you
put all 3 transparencies on top of each other, except that you flip the top one over so the numbers
are written backwards. This is suitable for an initial demonstration, or for a quick data collection
later, but if you want to keep the plot around for further discussion, you should transfer it onto a
single sheet with the digits all written in the normal fashion.
⇒ Note that for stem-and-leaf plots you retain the actual values of the numbers. The median
number of sit-ups in class 1 can be determined to be 45, while the median in class 2 is 50 sit-ups.
Exercise 5
[Link] Scatterplots and correlation
Scatterplots can be thought of as an extension of a two-way table for measurement of
(numeric) data that comes in the form of ordered pairs. An ordered pair consists of a pair of
numbers for one individual or object, such that the two numbers are two different measurements
on the same individual. Two examples might be height and weight of an individual, or
circumference and volume of a sphere. The data could be entered into a very large two-way table
with one unit of measurement for each category, but this would not give us a very informative
picture of the data. Instead, we can plot a point for each pair of data. The composite data set then
can be displayed as a scatterplot.
Example 6:
Suppose we have measured the height (in inches) and weight (in pounds) of students in a
class. A scatterplot of these data is shown below. Notice that the plot slants upwards as we move
from lower to higher heights. This is an example of a positive correlation.
We
could further emphasize this association by drawing a line that is as close as possible to all the
points; this line would have a positive slope (in more advanced classes formal procedures for
choosing this line are introduced, but for elementary and middle-school children a line “by eye” is
adequate). We might also emphasize the shape by drawing an ellipse around the majority of the
points. This ellipse would also slant upwards.
Example 7:
A second scatterplot might show number of hours of TV watched in a week vs. number of
book pages read. This might show a negative association, although with probably more scatter
than seen for the height/weight example. If we drew a line or an ellipse around these points, it
would slant downwards. Note that we would leave a few outliers outside the ellipse.
Finally, a third scatterplot might show no positive or negative association; in this case we
would say there is little or no correlation between the two variables. An ellipse drawn around these
points would not slant either upwards or downwards.
)
Interpreting the data:
As in all other cases, a good way to think about data interpretation is to look for patterns
and departures from patterns. In the example of height vs. weight, the overall pattern is that weight
increases with height. The line drawn through the points summarizes this increasing trend in much
the same way a mean or a median captures the center of a distribution plotted via, e.g., a line plot.
The particular departures from the pattern may also be interesting to consider. Is there
something truly unusual about these points, or are they just errors in plotting the data? Two
examples of unusual points are the single individual in the center of the weight distribution with
very low height, and the 3 points clustered to the far right and above of the rest of the points. Are
these data-measurement errors or are they really extreme points (outliers).
Initial preparation:
To make a scatterplot with data collected by students, start with either large sheets of quad-
ruled chart paper, or just tape a set of axes on the board or a piece of butcher paper. Find the
minimum and maximum values in the class for each variable. Mark off the axes in even intervals
to reflect these values, starting at or slightly below the minimum value. For very young children
stick with integers, but for older children the numbers can include non-integer scales as well. Label
the axes.
Exercise 6
A teacher surveyed her students about the amount of physical activity they get each week. She
then had their body mass index (BMI) measured. Use her data to make a scatter plot.
[Link] Box Plots
Box plots provide a visual representation of a five-number summary of data, consisting of
the median (the midpoint of the data range), the upper and lower quartiles (the numbers below the
highest quarter of the data and above the lowest quarter, respectively) and the largest and smallest
values (the extremes). Box plots are particularly useful for comparing distributions of the results
from several experimental conditions.
Because box plots are based on simple summaries, they can be used with fairly young
children certainly third graders, but even, in some cases, younger children. For the youngest
children, one can “lead up” to a box plot by using only the median (middle number) and the
whiskers to the extremes.
Box plots are an important type of graph to use with children. More than any other type of
graph, they focus the user on several key statistical concepts, perhaps the most important of which
is that the data can be summarized. Because box plots focus on representing summaries of the data,
the children are not distracted by issues such as gaps or multiple modes. In addition, the box plot
is a superb way of emphasizing and representing the variability inherent in real data in a way that
is computationally accessible to children. This is a key concept knowing that there is a way of
describing or summarizing variability is very important, and will lay the conceptual groundwork
for other summaries that can be introduced in high school. Finally, the box plot strongly
emphasizes the idea of the center of the distribution. Again, this is a key concept, especially as
children begin to compare results from different experimental situations.
Example 8:
Return to the original sit-ups/min. data set from the stem-and-leaf plot section. The five
number summary, written in the order of the lower extreme, lower quartile, median, upper quartile,
and upper extreme is:
34 41 45.5 50 64
A box plot of these numbers, plotted above the real number line for reference, would look like the
following:
Interpreting the data:
Half the scores will lie inside the box. Half the scores will lie outside the box on the
whiskers. Half the scores lie below the line in the box, and half lie above this line. One quarter of
the scores lie on each whisker and in each half of the box. Below are some additional examples,
without the number lines (and plotted on different scales), to illustrate possible shapes of resulting
distributions:
Exercise 7
1. Draw a box plot for the data set {3, 7, 8, 5, 12, 14, 21, 13, 18}.
2. Suppose that the box-and-whisker plots below represent quiz scores out of 25 points for
Quiz 1 and Quiz 2 for the same class. What do these box-and-whisker plots show about
how the class did on test #2 compared to test #1?
[Link] Time-Series Plots
Where a single individual or object has been measured for some characteristic at multiple
time points, a time-series plot is often used to show the results. Some examples where this type of
plot may be informative are: height of a child at different times over several years, air temperature
each morning over a month, or number of seeds sprouted each day in a model ecosystem over a
week. Time-series plots can be used in the youngest grades, as long as the units of measurement
fall into a range with which the children are comfortable.
Example 9:
Mary measured the height, in inches, of a marigold plant for 8 weeks from the time the
plant first came up. Her data:
week: 0 1 2 3 4 5 6 7 8
height (in): 0.25 0.75 1.5 3.0 5.5 7.75 9.5 10.5 10.75
When she plotted the results, they looked like the following graph:
Initial Preparation:
If you have measured one object as a class over time then you probably will make a single,
large graph using quad-ruled graph paper. Alternatively, if individuals or teams of children have
each measured an object or individual over time, then they will need to make individual graphs,
unless you want them to all plot their results on one large class graph. This should only be done if
there is only a limited number of such plots to be placed on the graph, e.g., about 6 or fewer.
Determine the smallest and largest measurement collected over the time of the study. These
will determine the scale of the dependent axis (the y-axis or vertical axis). In the data above, the
measured heights run from 0.25 inches to 10.75 inches; a reasonable scale for these data would be
½ inch for each square of ¼” graph paper (this will allow the graph to fit on one page, while
providing sufficient separation vertically to see the increase in height over time). On the dependent
axis, we have time of 0 to 8 weeks. We could use 4 squares/week for the time axis. Mark off the
time units on the horizontal axis 22 Version: July 2003 (for time-series plots, time is almost always
put on the horizontal axis). Mark off the measurement units on the vertical axis. Label the two axes
and add a title.
Graphing:
Plot the data, using dots, the same way as one would for scatterplots: find the point on the
horizontal axis first, and use a finger to trace upwards until the appropriate point on the vertical
axis is reached. Place a dot at this intersection. Connect the dots with straight lines.
Variations and extensions:
You can compare multiple plots by putting them in the same graph, using different colors
or line-styles to distinguish them. If you do this, remember to add a key. For example, one might
grow one marigold under each of three different conditions, and plot the growth of each plant on
the same graph.
It can be quite instructive as an exercise to plot the same data on two graphs that have the
same scale on one axis (e.g., the horizontal axis), but very different scales on the other axis. Try
doubling or halving the size of the units. Clearly, the two graphs represent the same data, but one’s
initial interpretation may be very different. If individuals in a class are allowed to choose their own
scales to plot data from a class data set this point will probably emerge naturally as you ask the
children to compare their plots. But you may need to guide the discussion of this issue, and to
suggest or request that some children replot their data on a scale which is the same as that used by
other children to make the point.
Exercise 8
Comparisons among Graph types
What is the distinction between bar plots, line plots, stem-and-leaf plots, and histograms?
When is it appropriate to use one but not another?
While each of these four graphing styles can be used to summarize integer counts of data,
each has its own advantages and disadvantages in representing different kinds of information. Bar
plots, commonly recommended in exercises aimed at introducing the concept of graphs to kids,
are really only appropriate for summarizing data sorted by a categorical (non-numerical) trait, such
as color (“blue” does not logically come before “green” and after “red”). For data which are
naturally classified by number, however, it would be more useful to use one of the other three
graphing styles. For instance, a line plot or a stem-and leaf plot is appropriate if your data set is
relatively small (~25-35) and consists of integer values. If your data is grouped by intervals of
numbers (i.e., 0-9, 10- 19, 20-29 etc.) or consists of non-integers (e.g., 4¾, 2.5), you should use a
histogram or a stem-and-leaf. A histogram is convenient if your data set is large, because you can
pool by larger intervals, and you do not need to plot individual points. A histogram may also be
useful when the range of data is very high, but the number of points is modest; by grouping data
into intervals, some picture of the shape of the distribution may emerge.
When should one use a box-plot instead of a histogram, line plot, or stem-and-leaf plot?
When you want to emphasize summary statistics associated with a data set (e.g., the median
and some measure of variability) then the box plot will do a better job at conveying this information
than will one of the other plots. In addition, when you want to compare several data sets, parallel
box plots are often a better choice than back-to back or separate histograms. In the context of
science experiments, where children are trying to make inferences about the effects of certain
experimental conditions, parallel box plots are more likely to provide a clear picture of a change
in distribution than are histograms. However, box plots do not allow you to see such aspects of the
distribution as where there are gaps or clusters, whether there are multiple peaks in the distribution,
or other such details. Box plots are also better for large rather than small data sets (e.g., more than
20 or so data points). When should you connect the dots on a plot (e.g., when should you use a
time-series plot instead of a scatterplot)? Generally, you should only connect the dots when the
horizontal axis represents time, and the points at the different times were measured on the same
subject. The line makes the implicit assumption that there was a measurement between the two
points being collected. If the data plotted at different time points are from different individuals,
then you should just represent the results as a scatterplot without connecting the dots. For example,
if one plots the diameter of a single evaporating circle of water over time (minutes), then one could
use a time-series plot. If Mary plots her results at 1 minute, Joseph plots his at 2 minutes, and so
on throughout the class, a scatterplot would be a better choice.
[Link] Dot Plots
When analysing data, you often need to study the characteristics of a single group of
numbers, observations, or measurements. You might want to know the center and the spread about
this central value. You might want to investigate extreme values (referred to as outliers) or study
the distribution or pattern of the data values. Several plots are available to allow you to study the
distribution. One such plot is the dot plot.
Dot plots are plots of points with the measured value on one axis and the category level on
the other axis.
A dot plot is constructed from a numeric variable. A second variable may be used to divide
the first variable into groups (e.g., age group or gender). In the two-factor procedure, a third
variable may be used to divide the groups into subgroups.
Exercise 9
In an airline training program, the students are given a test in which they are given a set of tasks
and the time it takes them to complete the tasks is measured. The following is a list of the time (in
seconds) for a group of new trainees.
61, 61, 64, 67, 70, 71, 71, 71, 72, 73, 74, 74, 75, 77, 79, 80, 81, 81, 83
Display the data in a dot plot.