Econbrowser's James Hamilton takes a detailed statistical look at the Graham et al. Lancet Vioxx study that was the subject of the now infamous cross-examination of Merck's CEO, where Mark Lanier argued (apparently successfully) that the concept of "statistical sigificance" is mere argle bargle (via Kirkendall). Economist Craig Newmark also has some interesting remarks. Economist/athlete Arthur de Vany, meanwhile, has some insight into the repeated mentions of Robert Ernst as a marathoner.
Lying with Statistics - Ernst v. Merck edition, Part II