Stan Young saw and sent me this article by Freedman in PLOS Biology on the consequences of unreproducible (bad) biological research.
Stan’s note (Stan is a PhD in Stats and Genetics):
Where experiments are run and the researcher can control everything,
well over 50% of claims fail to replicate. Environmental epidemiology
has to be much worse as nothing is controlled and the researcher usually
has an agenda and an eye on funding. They get the answer they want, they
publish. They don’t get the answer they want, they simply don’t publish.
It has a name, publication bias. In the medical world researchers have
to register their trials. You can count failed studies. In the epi world
you just move on.
Freedman et al. say bad science is very expensive. Well bad epi is much
worse as it is not the cost of the research, but the cost of the
corrective action that is so expensive.
I would agree.
The downstream economic effects of junk science in some cases include mistaken policy decisions and regulatory activity, for example.
The impact of the junk science claims is very severe in EPA policy making that creates compliance burdens in the billions, so the EPA sponsored research is exhibit one for this assertion.