Initiated-Event Model of Statistical Point Processes

This proposal seeks to introduce a new statistical methodology for event- history data to the mental health community. This model processes paired lists of event times to discern microscopic patterns in event timing, and is called the Initiated Event Model (IEM). The goals of the underlying model, and the output of the associated set of algorithms is a histogram of the various random delay times that would occur if events in one list helps precipitate or induce at least some of the "resultant" outcome events in the second list. The model continue to make useful predictions even when the initiating events occur so frequently that a given outcome could be due to any of the several proceeding initiating events. It also works in the presence of randomly occurring outcome events, uncorrelated with the set of proposed initiating events. The work for in the Phase I effort is designed to allow feasibility to be established. We propose to; (a) provide analytic and Monte Carlo estimates of the bias ed variance of the estimator; (b) investigate its domain of applicability; (c) redesign and rewrite existing-program code; and (d), (e) test its feasibility on two well-analyzed mental health data sets.