It should be so obvious as not to require restatement - much less a book - but the fact that correlation does not prove a cause-and-effect relationship between the datasets being correlated is widely, and often deliberately, ignored. From the correlation between serum cholesterol levels and deaths from coronary heart disease in the Framingham study, to time-shifted correlations between atmospheric CO2 levels and temperature, we are expected to accept correlation as proof of causation, despite all evidence to the contrary.
Enter the book, Spurious Correlations, by then Harvard student Tyler Vigen (tylervigen.com) (Hachette Books, 2015), to provide an entertaining collection of correlations - some of them very close - that nevertheless could not have any causal relationship. My favorite so far is Earnings per share of Domino’s Pizza Group vs. Economic loss due to Cybercrime, with a Pearson correlation coefficient of 98%.
It might be worth providing yourself with a copy of this little book, if only to brandish it the next time somebody says “it must be true, that correlation is almost perfect.” Unfortunately, its price in the used-book market ranges from $30 on up, much higher than the $2.50 that I spent for it in a used-book store here in the Philippines.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.