Stochastic models of the single case as the basis of clinical research design

The selection of a particular null hypothesis to be tested in a clinical situation, where only small numbers of homogeneous patients are available, often leads unwittingly to bias in its favor. It is frequently forgotten that failure to obtain a statistically significant result is equivalent to a nondecisive verdict, not to a verdict to reject a hypothesis (null or otherwise). Here the shortcomings of “extensive” models (in which the degrees of freedom do not exceed the number of patients in the sample) are discussed and a shift to an “intensive” model is indicated, based on the acceptance of a general underlying degree of variability in day-to-day behavior and affect.