I doubt it. Finding problems in e.g. a small-sample, 22-month study is one thing, finding problems in studies that cover entire nations over several decades is quite another.
Darrell Huff's style in How to Lie with Smoking Statistics was to marshal an army of minor quibbles to apparently undermine the entire body of research: find a small flaw in one study, a minor error in another, and pretty soon you've made it appear like all the research is flawed. He didn't need to actually prove any one study wrong.
He also relied pretty heavily on arguments that correlation is not causation because smokers are genetically predisposed to smoke, and may also be predisposed to cancer and other maladies. He even quoted R.A. Fisher's argument that people smoke to quell irritation and annoyance, and having a chronic inflammation or cancer might just be the kind of irritation that makes you subconsciously want to smoke.
No number of enormous well-conducted observational studies would prevent Huff from making up scenarios for other types of confounding. Check out Chapter 12 of the book.
point to 1 such study on smoking, where they did actual measurements on an entire nation. i bet you they just sampled a subset and called it statistically sound. ;D
Can't. I heard that when I was a conscientous objector at the university hospital in Trondheim, Norway. One of the physicians there told me that they'd done statistics on all patients. Norway has a single-payer system, so that would really be all patients. But it's a while ago and I can't find email addresses for anyone whose name I still remember, so I can't.
all patients and 'nation' seems a bit overconfident of said doctor (even though i respect their work a lot!)... i guess his world IS the hospital patients but vice versa i think the world or a nation includes (a lot) more than its hospital patients.
So I looked, starting at fhi.no, and found quickly found several studies that use data based on either the Norwegian cancer registry, the cause-of-death registry or both.
These aren't samples. The cause of death registry includes all deaths and the cancer registry includes everyone who's been diagnosed with cancer, even people who received no treatment. This means that studies that investigate e.g. cancer caused by smoking and use these data sources are nationwide rather than sampled.
