« So much for NY's registration of "assault rifles" | Main | Thoughts on the Confederate flag controversy »

More hits on the CT "study"

Over at Reason Online, Brian Doherty joins in the fun. I have an article, accepted and in edit, on how the grant-induced wave of medical studies (published in medical rather than criminology journals, so the editors and peer reviewers have no idea what they are dealing with) plays with the books. It's been a problem with medical articles in general, where the author sometimes has a vested interest in promoting some therapy or drug, and there the editors at least know what they are dealing with. One editor notes many different ways to cook the books, for example: run your study and use survival data from one, three, and five years out. If one and five years show no result, report only the results from three years, and never admit that you ran the other periods. Or ignore confounding variables (was this therapy only given to the less sick patients?). This "study" seems a clear example of those problems. It ends in 2005, for no convincing reason, and when extensions to 2010 or even 2014 would have found gun homicides rising. It compares CT, not to easily chosen controls (the region in which it is situated, or the entire nation) but to an artificial CT composed of parts of several States. With that sort of liberty, I'm sure you could have created an artificial CT that showed homicides went up, or down, or stayed the same.