Skip to Main Content

TC Library Online

A guide to library resources

Beware of spurious correlations

Sometimes statistical findings are linked so that it appears that two separate trends are related. Keep an eye out for this misleading technique, known as spurious correlation.

Check out this example from the site FiveThirtyEight. The blog Spurious Correlations, as the name suggests is full of even more examples.

Look out for dubious statistics. When something is an average is it the mean, median or mode? To get the mean, add each item's value and divide by the number of items. The median is the exact midpoint of a group. The mode is the most frequently occurring item in a group. Also look for missing specifics. How clearly are sample size and sample representation discussed? Is the wording vague, in a way that could be interpreted in several ways? Are the research methods clearly stated or conveniently left out? Correlation isn't causality. More stray cats are adopted during Shark Week does not mean Shark week causes cat adoptions. Watch out for misleading charts and graphs. Cutting off the bottom of a chart can inflate the appearance of change. Pie charts that show 100 percent but do not represent all options are misleading. Charts can also use size or scale of images to trick readers. Remember: if it seems unbelievable it probably is. Adapted from Huff, Darrell. How to Lie With Statistics. New Yor: W.W. Norton & Company, 1954.