Non-linear mixed-effects modeling for complex biopharmaceutical data

Abstract

Many phenomena in our world exhibit a nonlinear behavior, such as physiology, growth curves, etc. One might even consider that all phenomena that are apparently linearly, can only be considered as approximatively linear on a limited, though relevant, interval. The (generalized) linear mixed-effects model has the appealing feature of simplicity, which makes it very suitable for education and research, but the physiological interpretation of the model and the correctness of a simulation is often questionable at the limits of the apparently linear interval. Other phenomena lack completely the linear behavior, such as the case studies in this dissertation. It is clear from these examples that nonlinear mixed-effects modelling is a very useful technique deserving more attention by the statistical community, but the complexity of the methodology might be a hurdle. (Excerpt from Concluding Remarks)