On the Development of Clinical Statistical Systems for Psychiatry

The essential purposes of this presentation are to state briefly some of the basic properties of the data of clinical psychiatry and the implications that these properties have for the application of statistical methodology and inference to such data. One of the most important properties of data relevant to the study of psychopathological processes is that of a general underlying variability within the individual patient-state from one week to the next, or from one day to the next, or even from moment to moment, with respect to a large number of variables which are considered to be important in the description and characterization of mental illness. The realization of such variability as a fundamental property of psychiatric data leads to the conclusion that an observation of a patient-state at a single point in time or with respect to a short time interval cannot in general be expected to provide anywhere near as precise or statistically definitive a description of the patient-state as can be obtained from repeated observations of the same patient taken at relatively close intervals and over a period of sufficient duration. Thus if X represents a vector whose respective components are descriptive of various aspects of the patient-state as observed at a single point in time, then we are interested in estimating the distribution of X for each patient over an adequate period of time rather than in obtaining a single reading of X. This is with respect to obtaining an adequate description of the patient-state per se in statistical-clinical terms even before one uses such a description within a clinical experimental design such as might be used in the evaluation of two or more tranquilizers. The description of patient-states in terms of probabilities estimated from sequences of momentary or short-term states will in general include transition probabilities, i.e., probabilities that are dependent on the previous state. This concept of a stochastic definition of the pa-