45 pages • 1 hour read
Darrell HuffHow to Lie with Statistics
Nonfiction | Reference/Text Book | Adult | Published in 1954A modern alternative to SparkNotes and CliffsNotes, SuperSummary offers high-quality Study Guides with detailed chapter summaries and analysis of major themes, characters, and more.
Chapter 4Chapter Summaries & Analyses
Chapter 4 Summary: “Much Ado about Practically Nothing”
Chapter 4 focuses on creating impressions of a significant difference when the reality is small to nonexistent. Huff begins the chapter by discussing error in statistics and how its omission from reporting can lead to inaccuracies. He starts with the example of two people taking an IQ test. He takes issue with using the results as a measure of actual intelligence because the test considers only a few factors as indicating “intelligence,” but he focuses on the importance of the concept of error in interpreting the results. Huff explains how probable error works, the type of error he focuses on in the chapter over standard error. Because IQ test results have a range of statistical error, the results of students with an IQ of 98 and an IQ of 101 overlap. Thus, one should not be regarded as more intelligent than the other based on these numbers, because this difference places the students within the same statistical range based on the test’s “probable error” of +/- 3%. When looking at statistics with error, the reader should “keep that plus-or-minus in mind, even (or especially) when it is not stated” (59). If the probable or statistical error is ignored, the importance of the result becomes inflated.
In the next section, Huff says it is also essential to watch for differences that can be demonstrated but are too small to matter. He uses a test on cigarettes as an example. The results showed that the smoke’s content of nicotine and other chemicals was the same for all the brands, so “all the brands were virtually identical and […] it didn’t make any difference which one you smoked” (59). However, the company that appeared at the end of the list used the statistic to market its cigarettes as healthier than the others, even though this ranking was based on “negligible” differences that made it meaningless.
Chapter 4 Analysis
A short chapter in comparison to the preceding three, Chapter 4 serves as a closing for the book’s discussion of The Importance of Proper Sampling, particularly problems with data collection and analysis. Huff touches on statistical error, another feature he notes goes missing from many statistics. He explains the two types of error: probable error and standard error. He explains how to calculate probable error but only skates over the topic of standard error. While he admits that statisticians prefer standard error, probable error has more applicability regarding the IQ test used in his example.
Huff also returns to “little issues,” those differences that are so small that they aren’t statistically significant. For example, he uses a cigarette study that showed an insignificant difference between the contents of one brand versus another. This is the first of many examples in the book that focus on smoking. The health risks posed by smoking were only beginning to come into focus at the time, and the primary report on this topic wouldn’t be released for another decade.
