Slashdot Mirror
Meta-Research Debunks Medical Study Findings
jenningsthecat writes "From The Atlantic comes the story of John Ioannidis and his team of meta-researchers, who have studied the overall state of medical research and found it dangerously and widely lacking in trustworthiness. Even after filtering out the journalistic frippery and hyperbole, the story is pretty disturbing. Some points made in the article: even the most respected, widely accepted, peer-reviewed medical studies are all-too-often deeply flawed or outright wrong; when an error is brought to light and the conclusions publicly refuted, the erroneous conclusions often persist and are cited as valid for years, or even decades; scientists and researchers themselves regard peer review as providing 'only a minimal assurance of quality'; and these shortcomings apply to medical research across the board, not just to blatantly self-serving pharmaceutical industry studies. The article concludes by saying, 'Science is a noble endeavor, but it's also a low-yield endeavor ... I'm not sure that more than a very small percentage of medical research is ever likely to lead to major improvements in clinical outcomes and quality of life.' I've always been somewhat suspicious of research findings, but before this article I had no idea just how prevalent untrustworthy results were."
Gosh, it's a good thing that climate change studies are immune from this kind of thing! Otherwise there might be a bunch of people that are denigrating and insulting the skeptics for no reason!
"...history will look upon the act of depriving a whole nation of arms, as the blackest." --Ghandi
Medical people tend to understand statistics, to reuse an old saw, the way a drunk understands a lightpole--using it more for support than illumination.
/.ers know, for a statistical inference to be valid, the underlying dataset must be completely random. Not just sort of random, not just I'm pretty sure it's random, not just that's the best I could do random, not just they-said-it-was random. It must be completely random. Most of the time included random variables must also be completely independent (unless you're doing covariance studies, but let's not go there).
As most/all
Thing is, complete randomness and independence of variables within a human dataset is probably impossible, even in the big ones sponsored by NIH, the Census Bureau and so on. If that is so, then doing "statistical studies" on human datasets--which AFAIK is what the majority of medical studies attempt to do--is about as scientific as creationism.
So it's reasonable to criticize medical studies AND they realize that peer-review is, well, not so good AND they point out that erroneous conclusions persist....
Good thing AGW has avoided all that.
No brain, no pain.