It takes a family—a well-defined family—to underwrite familywise corrections

The false discovery rate (FDR) control procedures recommended by Matsunaga require the identification of a family of tests over which the procedures are applied. It is argued that Matsunaga's basis for grouping tests—that all the tests within the same null should be treated as a family, so long as there is a reasoning chain underlying the hypothesis—will, if applied in a principled (consistent) fashion, require bizarre and undesirable research practices. The underlying source of these difficulties appears to be an implicit (and unrealistic) focus on an isolated researcher conducting a single study, as opposed to a community of researchers with many studies and many tests.