How to Lie with Statistics

Chapter 8 deals with the issue of the post hoc fallacy and the problem of thinking correlation equals causation. Huff says this occurs when one of two items with a demonstrable correlation is assumed to cause the other. He describes the thought process as “If B follows A, then A has caused B” (89). However, this is not necessarily the case. Other factors might cause “B,” while “A” is not responsible for them.

Huff begins with a study on whether cigarette smokers get lower grades. He says the sample size and the significance of the correlation are good, but it needs to account for other explanations that could cause both smoking and low grades. Without a clear understanding of these unknown factors, conclusions about causation should not be drawn.

In the next section, Huff warns readers to carefully examine the relationship underlying any statistics to ensure they don’t fall for the fallacy. He then lays out the different types of correlation that can appear as causation. One is correlations made by chance, such as those found in small samples. Another is a correlation that has a real relationship but no way to tell which factor is the cause or the effect.