Monday, November 08, 2010

Dashing nerd and others say a large part of science is wrong

A genetics professor of mine used to tell his class, "Half of what I'm telling you is wrong, I just don't know which half."

I've heard variations on this sentiment frequently and fully believe them. I think there are many reasons that good research can be wrong, never mind the bad research.

John Ioannidis studies the phenomenon of inaccurate scientific conclusions. He is profiled this month in a readable article in the Atlantic (in which he is called a "dashing nerd"). Here are some excerpts from a scholarly article by John Ioannidis about identifying which science is most likely to be wrong.

The following characteristics predict that the research findings are less likely to be true:

  • Smaller studies
  • Smaller effect sizes
  • For a given set of data, more tests and fewer positive outcomes
  • Greater flexibility in designs, definitions, outcomes, and analytical modes in a scientific field
  • Greater financial and other interests and prejudices
  • Hotter scientific fields (with more scientific teams involved)
The article also uses the terrific term "data dredging" to talk about the phenomenon of trying elicit results that were not identified by the initial study design or analysis. Wikipedia calls it "the inappropriate (sometimes deliberately so) use of data mining to uncover misleading relationships in data."

via Derek Lowe

No comments: