Parameter Information Content During Model Identification Experiments

Abstract

In the identification of physiological systems it is often of practical interest knowing which time intervals during experiments provide most information on specific estimated model parameters. A common approach is based on visual inspection of the time courses of model output sensitivities determined for each parameter separately. This kind of analysis is however limited if strong correlation between parameter estimates is present. For this purpose, three approaches, based on different, yet related, statistical criteria are proposed in this study. Among the considered methods, a particular one provides a clear-cut partitioning of the experimental observation time interval for the various estimated model parameters. A case study is presented on the widely used minimal model of glucose disappearance and the so-called hot minimal model used to describe tracer glucose disappearance. The results demonstrate the usefulness of the proposed approach for defining the relevant time intervals for the identification of specific parameters, and for comparing different experimental protocols.