Quantitative studies of schizophrenic behavior

Behavioral data from eight psychiatric patients were collected and formed into 14 time series of 100 weekly rates (percentages of positive entries for frequent behaviors such as pacing, talking, hand-movement, etc.). A variety of methods of time-series analysis were applied to determine optimal predictive schemes in an effort to develop an inductive approach to input-output behavioral modelling. Some attempts at numerical prediction into new data were successful, and the results suggested that the best overall approach is likely to be nonlinear modification of auto-regressive schemes. Nonlinearity and nonstationarity of such time series are prominent features, but deterministic trends appear strong enough to justify extensive work. Application of artificial inputs to some of the models suggests that protracted oscillatory responses to stimulation may be characteristic for some schizophrenic patients.