*Quality Digest,* March 5, 2018

Manuscript 328

www.spcpress.com/pdf/DJW328.pdf

1

March 2018

The

Empirical Rule
What the Average and Standard Deviation

Tell You About Your Histogram

Donald J. Wheeler

How can we use descriptive statistics to characterize our data? When I was teaching at the

University of Tennessee I found a curious statement in a textbook that offered a practical answer

to this question. This statement was labeled as “the Empirical Rule,” and it is the subject of what

follows.

THE EMPIRICAL RULE

While a statistic may provide a mathematical summary for the data, it has to be

understandable before it can truly be said to be descriptive. While the average is easy to

understand, most students have trouble understanding the standard deviation statistic. The

empirical rule converts the average and standard deviation statistics into comprehensible

statements about the data using three intervals centered on the average. The first interval has a

radius equal to the standard deviation statistic, the second has a radius equal to twice the

standard deviation statistic, and the third has a radius equal to three times the standard deviation

statistic. The three parts of the empirical rule are:

Part One: Roughly 60 percent to 75 percent of the data will be found within the interval

defined by the average plus or minus the standard deviation statistic.

Part Two: Usually 90 percent to 98 percent of the data will be found within the interval

defined by the average plus or minus two standard deviations.

Part Three: Approximately 99 percent to 100 percent of the data will be found within the

interval defined by the average plus or minus three standard deviations.

“But can it really be this simple?” “Don’t we need to assume that our data are described by

some particular probability model before we can compute such specific percentages?” In what

follows I will attempt to answer these questions by looking at several data sets and then

explaining the source of this guide for practice.

We begin with the wire length data. These 100 values have an average of 109.19 and a

standard deviation statistic of 2.86. As shown in Figure 1, the three intervals of the empirical rule

contain, respectively, 69 percent, 95 percent, and 100 percent of the data.

*Donald J. Wheeler*
*The Empirical Rule*
www.spcpress.com/pdf/DJW328.pdf

2

March 2018

100

105

110

115

120

109.19

2.86

2.86

2.86

2.86

2.86

2.86

69%

95%

100%

100.6

117.8