EcoSta 2018: Registration
View Submission - EcoSta2018
Title: Estimation and inference of time-varying coefficients in non-linear ordinary differential equation models Authors:  Naisyin Wang - Univ of Michigan (United States) [presenting]
Abstract: The use of ordinary differential equations (ODEs) in modeling dynamic systems has gained high popularity in the recent decade. The physiological meanings of ODE parameters are often useful in enabling scientists to gain better understanding of the underlying system. On this regard, both estimation and inference procedures are essential. Even though various time-varying coefficient ODE model has been considered previously. The inference procedures considered earlier tends to be similar to what has been used in parametric models. We propose a new set of estimation and inference procedures for time-varying ODE coefficients. Our methods take into account features that are unique for ODE estimation and, as such, are adaptive in nature. The validity of the proposed procedures is justified through asymptotic properties. The numerical efficacy of the methodologies is illustrated using both synthetic and real-world data-sets.