Statisticians can prove almost anything, a new study finds

JunkScience.com has been saying (and proving) this for 15+ years.

The National Post reports:

Catchy headlines about the latest counter-intuitive discovery in human psychology have a special place in journalism, offering a quirky distraction from the horrors of war and crime, the tedium of politics and the drudgery of economics.

But even as readers smirk over the latest gee whizzery about human nature, it is generally assumed that behind the headlines, in the peer-reviewed pages of academia, most scientists are engaged in sober analysis of rigorously gathered data, and that this leads them reliably to the truth.

Not so, says a new report in the journal Psychological Science, which claims to show “how unacceptably easy it is to accumulate (and report) statistically significant evidence for a false hypothesis.”

Read the National Post article.

2 thoughts on “Statisticians can prove almost anything, a new study finds”

  1. I think that no one can prove anything with statistics. You only can fail to reject the null hypothesis, which leaves you with an association, not proof of a cause and effect. To get to that you must find some physical mechanism that is a path between A and B. The kinds of studies that these guys did can never provide anything more interesting that cocktail party conversation.

Comments are closed.