Reply #48: Its not always bad to make an inference based on a statistical correlation.

[D23MIURG23]

Such correlations are often the basis of hypotheses that are investigated more thoroughly to useful ends. The issue is that without additional lines of evidence and really diligent exclusion of variables such conclusions are highly questionable.

Double blind study design is also not uniformly essential, but for certain questions (Drug toxicity and efficacy etc.) it is considered indispensable and standard practice. Generally though, you will find it more in health sciences than anywhere else.

The real problem here, as you allude to, is the way the media tends to handle the business of scientific reporting. They tend to over hype studies that will produce interest regardless of the strength and acceptance level of the study (see "cellphones cause brain tumors", "cannabis causes schizophrenia" etc.) while misrepresenting levels of controversy surrounding well established ideas (see "evolution", "global warming"). Other times they allow voices weigh in on scientific topics who have no business discussing anything more complicated than basic addition and subtraction (see "Jenny McCarthy"). Personally I blame the never ending quest for ratings. Science is in the category of things many people consider dry, so the impulse is to "sex it up" with various bits of controversy, celebrities, and other things that Americans like, which in the end tend to undermine the quality of the information presented.