Autoregressive models for describing non-linear changes in biological parameters fitted using bugs

Many biological processes give outcome data which show a curvilinear associ ation with time which tends to an asymptote. We show how autoregressive mod els can be used to describe this association within individual subjects. We also present a Bayesian approach implemented using statistical software, B UGS, to fit these models in a multi-level (hierarchical) setting that descr ibes variation in the association between subjects. Peak expiratory how dat a from a clinical trial involving subjects with asthma are used to illustra te the methods. Copyright (C) 1999 John Wiley & Sons, Ltd.