Correlation and Causality (A)
In many instances, the knowledge you've gained from our course puts you in a position where you are better able to DECIPHER and CRITIQUE research summaries that appear in the newspaper. Once in a blue moon, however, it works the other way around: newspaper articles provide insight into and/or reinforcement for important elements of our course content. A good example of this latter situation comes from today's local newspaper.
On page A-8, there's an article entitled ''Study: The More Hair You Lose, the More Your Heart's at Risk." The second sentence of this article says this:
A study from Boston's Brigham and Women's found a clear association between the degree of hair loss and increased risk of heart disease: the less hair, the more risk.
Sounds like a correlational study, doesn't it? And do you remember the "warning" about correlation and causality that appeared near the end of Chapter 3 of our text? Here's a portion of what appears on page 77 of READING STATISTICS AND RESEARCH:
It's important for you to know that a correlation coefficient does not speak to the issue of cause-and-effect. In other words, whether a particular variable has a causal impact on a different variable cannot be determined by measuring the two variables simultaneously and then correlating the two sets of data.
Now consider another passage out of today's newspaper article. In my opinion, these next 4 sentences do a truly wonderful job of showing how two measured variables can sometimes turn out to be correlated NOT because one of those variables "causes" the other BUT RATHER because both of the variables are connected to a third unmeasured variable . . . and it's the third variable (that didn't get measured that's really the driving, causal force behind EACH of those first two variables that were measured and found to be correlated.
An author of the study cautioned that while the results showed a link between hair loss and heart disease, "it clearly has no public health impact for the public or physicians now." That's because the hair loss is really a marker for an unknown cause for such an association. In other words, said A. Ajani, an associate epidemiologist at the hospital, a loss of hair is a surrogate for something else that actually may be raising the risk [of heart disease]. Because hair loss is believed to be tied to levels of male hormones, it suggests, he said, that the level of such hormones or balance between them in men with hair loss may have a more direct influence on the increased risk of heart disease.
What a wonderful example of how two variables may be correlated not because of a causal link BETWEEN them but rather because each of those two variables is causally linked to a ''THIRD VARIABLE."
Copyright © 2012
Schuyler W. Huck