Diminished by Misuse: Statistics In an Alien World

Statisticians aren’t liars... just misunderstood

Published: Monday, December 12, 2016 - 11:07

Statistics has gotten a bad rap. People love to quote Mark Twain (“There are lies, damn lies, and statistics,” alternatively attributed to Benjamin Disraeli), Vin Scully (“Statistics are used much like a drunk uses a lamppost: for support, not illumination”), or Stephen Leacock (“In ancient times they had no statistics so they had to fall back on lies”).

For statisticians, these jokes have become quite tedious. Avoiding small talk at cocktail parties where quips are likely to come up or lying about one’s profession (“I’m a kind of mathematician” sometimes works) are not really satisfying alternatives to the lines that people have saved to shower on the innocent professional. What’s a statistician to do?

Unfortunately, since the application of statistics is indeed frequently misunderstood or misused, many people’s perceptions of statisticians are colored by their certainty that statistics represents a specious approach at best, and that statisticians are mere liars.

As Tyler Vigen’s book Spurious Correlations (Hachette Books, 2015) points out, statistics are indeed habitually misused, especially when it comes to understanding correlation. The statistician’s mantra, “Correlation does not equal causation,” falls on the deaf ears of those who insist that correlation does indeed equal causation. Charts show this, don’t they? Vigen shows the dramatic correlation between rates of margarine use, for example, and divorces per 1,000 people (see figure 1). These rates have nearly identical paths, with a correlation of 98.9 percent.

Figure 1: Correlation between margarine use and divorces

Clearly, the ups and downs of margarine use have nothing to do with divorce rates—that’s common sense, except maybe in some fantasy best seller. Both statistics may be accurate, but drawing a conclusion that they correlate is an irresponsible use of statistics.

To discover correlation of factors, scatter diagrams are a useful way to begin to analyze data, but cannot suggest specific causal relationships that demand further analysis. In a real estate example, two scatter diagrams indicate correlations between square feet and selling price of a home (see figure 2), and the number of bedrooms and selling price (see figure 3).

Figure 2: Scatter diagram for square feet and selling price of a home

Figure 3: Scatter diagram for number of bedrooms and selling price

One might see that the influence of square footage on selling price seems to be clear, since the slope of the line slopes up and to the right. The influence of the number of bedrooms is not quite as clear: There is a correlation, though, demonstrated by an upward slope of a trend line. One point that must be made is that these two factors—number of bedrooms and amount of square footage—are a least in the same context, and the attributes being studied are from the same house, so it is clear that they are related. The problem comes when causation is attributed to unrelated factors. Data for to the number of films Nicholas Cage has made and the number of swimming pool deaths may reflect similar or even exact patterns, but there can be no assumption of correlation or causality, unless one is into wild conspiracy theories.

Statistical analysis yields critical information that supports data-driven decision making. But this analysis, like a surgeon’s knife, must be applied carefully and with the skill to understand its proper application to available data. It’s as simple as that. Statistics has gotten a bad rap because of malpractice—and not by statisticians.

About The Author

Barbara A. Cleary

Barbara A. Cleary, Ph.D., is a teacher at The Miami Valley School, an independent school in Dayton, Ohio, and has served on the board of education in Centerville, Ohio, for eight years—three years as president. She is corporate vice president of PQ Systems Inc., an international firm specializing in theory, process, and quality management. She holds a masters degree and a doctorate in English from the University of Nebraska. Cleary is author and co-author of five books on inspiring classroom learning in elementary schools using quality tools and techniques (i.e., cause and effect, continuous improvement, fishbone diagram, histogram, Pareto chart, root cause analysis, variation, etc.), and how to think through problems and use data effectively. She is a published poet and a writer of many articles in professional journals and magazines including CalLab, English Journal, Quality Progress, and Quality Digest.