Fitting time-dependent multicompartment models: a case study.

A multicompartment model with time-dependent transfer rates is fitted to data on ovarian follicle dynamics in mice. The fitted model arises from an interplay between parametric and nonparametric approaches to fitting curves to these data. Nonparametric regression estimates, in the form of spline smoothers, are used in conjunction with the biologically meaningful general class of multicompartment models to suggest refinements to the parametric model. One result of this interplay is the suggestion of using three-stage step functions for the compartmental transition rates; the resulting curves mimic closely the nonparametric regression estimates.