|
Abstract:
|
We focus on the evaluation and optimization of designs in the context of nonlinear mixed-effects models. We present our methodology based on an approximation of the expression of the Fisher information matrix for these models, using a first-order linearization of the model around the random effects. Models including additional parameters for covariates and/or inter-occasions variability also are considered. Based on this methodology, we present PFIM and PFIMOPT, R/SPlus functions for population designs evaluation and optimization, respectively. Optimization uses the D-optimality criterion. Two algorithms are implemented, a general Simplex algorithm and a more specific one, the Fedorov-Wynn algorithm. Models can be supplied either analytically or under differential equations form. We illustrate the usefulness of PFIM and PFIMOPT on population pharmacokinetics/pharmacodynamics examples.
|
