Here's an update to some effect size content in the textbook.
Figure S1.12
First, in Supplemental Chapter S1, there's a comparison of psychological to medical effect sizes (page 477 of the 3rd Edition). Figure S1.12 reproduces a chart by Bushman and Anderson (2001) comparing the correlation between violent media and aggression to effect sizes from various medical studies, such as the association between smoking and lung cancer. The chart makes psychology look good--our effect sizes appear larger than common medical wisdom!
I've learned that this figure is problematic. One issue with the figure concerns its medical study comparators. The examples suggest that effect sizes for medical treatments and risks are rather small (such as the link between asbestos and cancer or lead exposure and IQ). But over the years, critics have argued that the examples in Bushman and Anderson's figure are artificially low or inaccurately computed. (Notably, Bushman and Anderson have defended their figure, more than once.)
Another issue is that the effect sizes for violent media and aggression are based on meta-analytic estimates that vary depending on the techniques used. Certain researchers report medium sized effects, but others argue that media effects on violence are negligible. (This debate led me to remove a media violence meta-analysis from Chapter 14 when I prepared the 3e.)
Ultimately, I am persuaded by Ferguson's (2009) explanation of how effect sizes of medical treatment studies vary dramatically, depending on whether they include "hypothesis irrelevant" cases. Especially when effect sizes are based on population-based epidemiological frequency counts, they can drastically underestimate the effect size of a vaccine, drug, or risk factor.
Chapter 8: Aspirin and heart attack
Chapter 8 uses the classic example of "aspirin prevents heart attack" as an illustration of how a very small effect size (r = .03) can, in fact, have large life-or-death impact. But if the aspirin-heart attack effect size came from a study that included hypothesis-irrelevant cases, it probably underestimated the true effect size. Indeed, the Physician's Health Study estimated that aspirin reduces the risk of a first heart attack by 44%--a much larger effect.
As always, I am grateful to colleagues who read my text carefully and alert me to problematic examples. Thanks!
