Understanding Data Summaries
Learn how measures of center and spread describe numerical data, and how to choose and communicate summaries that fit the data.
Three Ways to Describe the
A measure of describes a typical or central value, but different measures capture in different ways.
: Add all the numerical values and divide by how many values there are. Because the uses every value, an unusually high or low value can pull it toward one end of the data.
: First order the values. The is the middle value; if there are an even number of values, it is the of the two middle values. It depends on position rather than the size of every value, so it is less affected by extremes.
: Find the value that occurs most often. There may be one , multiple modes, or no if all values occur equally often. Unlike and , can also describe categories—for example, the most common preferred color.
Worked example: For the ordered values , there are five observations, so the middle one is 5. The is also 5 because it occurs twice, more often than any other value. For the , add all five values and divide by five:
The high value 20 pulls the above the . The and both identify 5, but they do so for different reasons: 5 is the middle value and the most frequent value. As a hypothetical change, if 20 were replaced by 30, the would become , while the and would still be 5. These measures describe different features and need not be equal.
Measuring the Span
The describes the full span of a numerical data set, from its smallest to its largest value:
For , the maximum is 20 and the minimum is 4, so the is . The values extend from 4 to 20, a span of 16 units.
is quick to calculate, but it uses only the two endpoints. An unusually large or small value can make it much bigger. Also, the does not show how values between the endpoints are distributed: it gives the span, not the pattern of values across that span.
Choosing and Reporting a Useful Summary
Choose a summary that fits both the kind of data and what you want to communicate.
For numerical data without strong outliers, the can be a useful because it incorporates every value. Pair it with a measure of , such as the .
When numerical data include an extreme value or are unevenly distributed, the often gives a more representative . The minimum and maximum show the span, but remember that the is also influenced by extremes.
For categorical data, such as colors or types of transport, report the or a frequency count. , , and do not describe categories.
Interpret measures together rather than in isolation. For the five values , a summary could say: “There are five values; the is 5, the is 16, and the values run from 4 to 20.” This gives the reader the number of observations, a measure of , and the span. A frequency summary can report how often each category—or each numerical value—occurs. Matching the summary to the data helps explain what the measures do—and do not—show.