Wednesday, February 29, 2012

Biased Samples in Scientific Journals

If you get positive results, then I have one major talent, and that is getting published in mainstream journals.
—Yudhijit Bhattacharjee quoting psychologist Daryl Bem.  March, 2012.  “Paranormal Psychologist.”  Discover magazine.  Bem was talking to Charles Honorton, who sought to verify the existence of ESP’s like telepathy.  Note: This illustrates why statistics reported in scientific journals are biased.  If 100 studies are conducted on the ability of people to exhibit telepathy and only two suggest telepathy exists, those are the only two studies that will be published because they are the only ones (out of the 100 studies) that are interesting.  In reality, the probability of telepathy existing is 2%, but according to the journals, it will be 100%.  

Perhaps this explains the following quote.

If I had to guess whether Wikipedia or the median refereed journal article on economics was more likely to be true, after a not so long think I would opt for Wikipedia.
—Tyler Cowen