Sunday, March 13, 2016
Medical journal editors make fools of themselves
The major British medical journals are heavily politicized -- both Lancet and BMJ. Lancet even criticized the Iraq intervention under George Bush II. When they stick to their knitting -- medical research -- they do publish some good studies and are prestigious because of that. But the people running the journals are obviously Left-leaning -- like most academics -- so they can't help misusing the platform they have to hand in order to promote their Leftist views.
That is of course bound to be an amusing exercise. Leftist claims are so counter-factual that a political study is bound to require all sorts of distortions and evasions to make any case at all. And the political studies they do publish are so unscholarly that the editors have obviously put their brain into neutral before publishing them.
The latest such study in Lancet ("Firearm legislation and firearm mortality in the USA: a cross-sectional, state-level study") is a good example of that. Dean has put up some good comments on it but I would like to point out a few things too. The article purports to show that some gun control laws do reduce deaths from guns. But the methodology behind the article is so naive that no such study would normally get published in a good academic journal.
What the researchers did was to look at gun deaths in the various States of the USA and compare the death rates with the various gun control laws in the various states. And they found that States with strict laws (e.g. Massachusetts) had fewer deaths per head than did States with more relaxed laws (e.g. Louisiana). And from that they concluded that certain gun control laws should be implemented nationwide. And if that were done gun deaths would drop dramatically nationwide.
Can anybody see something wrong with that reasoning? I am inclined to think that anyone with an IQ over 100 could. The study is an example of a fallacy that is all too common in the medical research literature: The fallacy that correlation is causation. If you ever do a course in logic, one of the first things you will be told is that correlation is NOT causation. To prove causation you need a strictly controlled before-and-after study.
To show how fallacious the reasoning is in the Lancet study, let me cautiously suggest that there are other factors that could lie behind the correlation between gun control and gun deaths. Let me suggest, for instance, entirely hypothetically, of course, that Massachusetts might have fewer gun deaths, not because of its laws but because they have fewer people there from a certain population segment that is more prone to gun deaths and crime generally. I am not going to nominate the population segment concerned because that does not affect the argument. But if the thought "Massachusetts has fewer blacks" pops into your mind, who am I to correct you? Entirely as a matter of academic interest of course, Mass. is less than 2% black and Louisiana is about a third black.
The study might have been of some interest if it had controlled for other factors. So I read the study carefully looking for that. But in fact they controlled for unemployment rate only. Control for race would be politically incorrect, of course. All men are equal, don't you know? It would be a compliment to call the study sophomoric. It is a disgrace to the journal. By normal scientific standards, it should never have been published.