Calculation-heavy courses teach t-tests, ANOVAs, and other statistical procedures. But students don't learn to spot common errors in statistical statements -- a problem when they need to read statistical results and interpret them for their own lives, said Utts.
In a recent paper published in the journal American Statistician, Utts describes seven common ways that statistics are misinterpreted, in news reports and elsewhere. The most insidious mistake, she said, is confusion about cause-and-effect relationships.
"Often, an observational study will link two variables where you'd like to think that there's a cause-and-effect relationship," said Utts. But based on the way these studies are done, there is no justification for concluding that cause-and-effect relationships exist, she said.
For example, a recent study found higher suicide rates among women who have had breast implants. "People might like to conclude that breast implants are causing suicides," Utts said. But the study design doesn't support that result. To avoid confusion when interpreting observational studies, you have to include possible alternate explanations, Utts said.
"For instance, women's self image could confound the study on breast implants and suicide," she said.
Confusing the terms "average" and "normal" is another common pitfall. "We're always hearing that 'the normal temperature for today is 80 degrees,'" Utts said. "But that's the average of temperatures recorded on this date in the past." Including a range of likely values with reported averages -- for example, "temperatures on this date have varied from 64 to 91 degrees" -- is more useful
'"/>
Contact: Andy Fell
ahfell@ucdavis.edu
530-752-4533
University of California - Davis
20-Oct-2003