A simulation study of parameter estimation in the one and two compartment models

Abstract

Few attempts have been made to examine the statistical problems that the user of compartmental models must face. Some properties of the estimators of parameters for one and two compartmental models based on nonlinear estimation were studied through simulation. Of particular interest were the effect of the experimental design and the effect of different error structures on the empirical sampling distribution for the estimators. For the one compartment model it was found that nonlinear estimation yielded essentially unbiased estimators that were normally distributed unless the random error for the model was large. In the two compartment model simulations, bias appeared in the estimators to the extent that bimodal sampling distributions of the estimators were observed as the random error for the model was increased.