Design of Informative Identification Experiments for Behavioral Interventions

This paper describes the design of informative system identification experiments for behavioral interventions based on Social Cognitive Theory (SCT); prior work has shown that these can be expressed as MIMO dynamical systems. The design methodology relies on identification test monitoring to determine the shortest experimental duration resulting in a model estimate with a prescribed level of accuracy. This is accomplished by generating statistical uncertainty regions that are the result of nonparametric transfer function estimates using periodic (and orthogonal in frequency) multisinusoidal inputs. The procedure is illustrated with an example of a simulated behavioral intervention for promoting physical activity.