Risk Assessment of Complex Evolving Systems Involving Multiple Inputs

Abstract

When monitoring complex evolving systems a question that often arises is to detect causal chains of events. Do particular inputs force the system to produce new events or prohibits them? This can be also considered as a risk assessment for the systems response. In this work we present two methods of estimating the effect of multiple inputs on a complex neurophysiological system. Both the response and the stimuli are very long binary time series. The first approach is a non-parametric one and describes the linear and the non-linear association of the stationary point processes by estimating the second- and third-order cumulant density functions. The second approach is a parametric one and the association between the inputs and the response of the system is described by a logistic regression model which takes into account the system’s internal processes.