Parameter estimation for compartmental mixture models

Abstract

Compartmental models are widely used in many applications, including pharmacokinetics, phisiology, sociology, and ecosystem modeling. Stochasticity may be introduced into a compartment system in many ways. In some instances, the transition rate coefficient is random, either between replicates, or between particles within a replicate, introducing stochastic mixture models. The error structure for some one and two compartment mixture models is derived, based on the randomness of the transition rate. The effects of using the model variances derived in the estimation process are investigated. The effect of model misspecification in the parameter estimates is also investigated, using several data sets and simulations.