There are no substitutes for p values when a statistical interrogation must lead to a decision. It will therefore come as no surprise when I say that their use requires careful, deliberate thought. Even in the simplest of experiments, investigators must think critically and prospectively about the degree of community protection they wish to provide as they determine the type I and type II errors. As I pointed out in chapters 7 and 8, the application of p values to experiments in health care is even more complicated when these experiments have more than two treatment arms and multiple endpoints. In all cases, the experiment must be executed concordantly, according to the protocol, so that the resulting p values are interpretable. These endeavors are not to be undertaken lightly.
KeywordsPosterior Distribution Loss Function Prior Distribution Bayesian Approach Asthma Prevalence
Unable to display preview. Download preview PDF.