In a recent interview, Gary Taubes calls picking data that support your conclusion “Bing Crosby science.” This comes from a song by Bing Crosby that begins “You’ve got to accentuate the positive, eliminate the negative.”
Taubes uses the phrase to refer specifically to epidemiology, though it applies to all science. He credits “a Scottish researcher” with coining the phrase, but doesn’t say any more about who this researcher was.
Related post: Amputating reality
We need someone write a song about best linear unbiased estimators and uninformative priors. And it needs to be REALLY catchy! ;-)
For an eye-opening exposé of an entire field practicing Crosby science, see the discussion of modern epidemiology at http://www.mbswonline.com/upload/presentation_Stan5-28-2009-14-28-24.pdf
Summary: positive results get published, even if false. The field has evolved to a standard of evidence that maximizes the probability of finding something publishable, even as reproducibility of results goes away.