—The essay on P values was meant to alert readers to possible adverse numeric consequences of repeated, mindless, automatic grinding out of statistical tests. (This kind of unplanned data dredging might employ an automaton named Robostat.) I also wanted to point out the usefulness of the-binomial expansion [(p + q)n = 1.0] in some ordinary biomedical situations.
Dr Coulter raises an interesting technical issue in hypothesis testing. Hypothesis formation and testing are somewhat different, and much more complex, subjects than those addressed in the essay on P values. Research hypotheses involve elements that are much more "scientific" than mere probability phenomena. Researchers incorporate many elements into their hypotheses, elements such as prior experience, extrinsic knowledge, biologic plausibility, subjective judgments, and so on.
For example, the essay on P values omitted discussion of one-tailed vs two-tailed statistical tests. The decision whether to use a one-tailed or a two-tailed test depends on the
BROWN GW. More on the P Value-Reply. Am J Dis Child. 1991;145(3):249–250. doi:10.1001/archpedi.1991.02160030013004
Pediatrics in JAMA: Read the Latest
Customize your JAMA Network experience by selecting one or more topics from the list below.