[mttpgn]

An excerpt from _How to Lie with Statistics_ by Darrell Huff (1954):

> Take this one: "27 percent of a large sample of eminent physicians smoke Throaties--more than any other brand." The figure itself may be phony, of course, in any of several ways, but that really doesn't make any difference. The only answer to a figure so irrelevant is "So what?" With all proper respect toward the medical profession, do doctors know any more about tobacco brands than you do? Do they have any inside information that permits them to choose the least harmful among cigarettes? Of course they don't, and your doctor would be the first to say so. Yet that "27 percent" somehow manages to sound as if it meant something.

That book specifies many other examples (from this time period in America) of misleading claims that sound statistically significant upon an uncritical, cursory reading.

[capnrefsmmat]

Ironically, Huff was then hired by the tobacco industry to write a book undermining the evidence that smoking causes cancer. It was to be called How to Lie with Smoking Statistics, but never got published: https://www.refsmmat.com/files/papers/huff.pdf

[croemer]

I think you don't mean "statistically significant" here, but something like "relevant" instead. Something can be statistically significant and entirely irrelevant if the effect size is too small.