News for hackers. Stuff that matters.
Re:Statistics Done Wrong – The woefully complete guide by davmre on Tuesday November 05, 2013 @02:25PM (#6677467)
Something like that idea has actually already been the inspiration for at least one startup: MetaMed (http://en.wikipedia.org/wiki/MetaMed, http://nymag.com/health/bestdoctors/2013/metamed-personalize...) does meta-level analysis of the medical literature to determine which treatments seem effective for rare conditions, taking into account the sample size, statistical methodology, funding sources, etc. of each study.

Of course, medicine might be unique as a domain in which individuals are willing to pay vast sums of money to obtain slightly more trustworthy research conclusions, and the profit motive has obvious conflicts with "benefit to humanity" (if someone pays you to research a treatment for their disease, do you post the findings when done? Or hold them privately for the next person with the same problem?). But maybe there are other domains in which the market could support a (non-billionaire's) project for better-validated research.

Re:Statistics Done Wrong – The woefully complete guide by gwern on Tuesday November 05, 2013 @03:34PM (#6678045)
MetaMed, as far as I know, does basically just customized literature reviews; it isn't doing anything I'd recognize as 'meta-level analysis' like the work done by the Cochrane Collaboration or using meta-analytic techniques to directly estimate the reliability of existing medical treatments or beliefs.