Sir.—I want to bring to your readers' attention the problem of type II errors in clinical research studies that reach conclusions of "no difference." Please forgive me as I cover some territory that is no doubt familiar, but I want to develop my argument fully.
An investigator typically analyzes data by subjecting them to a statistical test, to see whether the difference observed is likely to have resulted from chance alone. When P<.05, the investigator concludes that there is a less than 5% chance that a difference of this magnitude could have resulted from chance, and so he or she concludes that the difference is real. So far, so good. But what happens when P>.05? Many investigators unfortunately conclude, without further analysis, that the two outcomes studied are not different. What's worse, many medical journals let these investigators publish this erroneous conclusion. Merely knowing that P>.05 does not tell