Cherry-picking, statistical adjustment, and fishing expeditions
December 5th, 2011
My Jockology column in this week’s Globe and Mail takes a look at three questions:
- How do cigarettes help marathoners run faster?
- Why does eating red meat cause car crashes?
- Does caffeine cause breast cancer?
In each case, I analyze studies that seem to “prove” these surprising findings, and identify the errors (cherry-picking data, inadequate statistical adjustment, and fishing expeditions) that lead to these conclusions. Basically, it’s a “how to assess medical research” primer. Read the whole thing here.
(Hat tip to Travis Saunders for his blog post at Obesity Panacea about the smoking study.)
Great article. Here’s a great cartoon illustrating “fishing expeditions” (and the role of “science” journalism):
http://xkcd.com/882/
@jb: Ha! Awesome — I recommend everyone click on the xkcd link for a good laugh (and concise demonstration of the problem with fishing expeditions).
It’s covered fairly well in “bad science” in simple laymans terms so even I can understand.