International Biometric Society Clustering and Partitioning of Functional Data for Understanding the Heterogeneity in Psychiatric Diagnosis

In the past two centuries progress has been made in many medical areas with the develop-ment of objective tests and understanding of the cause (etiology) and mechanism of diseases (pathophysiology). Unfortunately, the attempts to understand causes of psychiatric illness by studying genetic or brain functions did not yield the hoped for results and psychiatry still de-pends on non-objective descriptions of syndromes. Some psychotropic drugs can specifically induce remission, suggesting an experimental approach to understanding the pathophysiologi-cal processes. This approach requires evidence that the medication directly affects the disease mechanism, rather than exerting benefit by non-specific compensation. For each particular pa-tient who remits during drug treatment, it is unknown if s/he benefited from a specific pharma-cological action or from a non-specific placebo effect or spontaneous remission. If psychiatric illness consists of distinct diagnostic categories (e.g., subjects who’s illness has a mechanism that can be affected by a specific pharmacologic agent, vs. those illness is not affected by this pharmacologic agent), then we should be able to cluster responders to the treatment into distinct classes based on, for example, course of the severity of symptoms over time, med-ical history and other measures. This would allow to isolate a purer subset of the subject