Presentation: Designing Studies for ODE Parameter Estimation
Speaker: Chloe Krakauer, Graduate Student, UW Biostatistics
Abstract: Systems of ordinary differential equations (ODEs), which formulate the rate of change of variables over time, are a useful tool to describe expected trajectories of variables in a variety of disciplines. Across many disciplines, including pharmacokinetics and immunology, studies are conducted to estimate unknown values of parameters in these systems. Resources are often limited, particularly number and timing of samples, which can severely impact efficiency of parameter estimation. Many systems do not have analytic solutions, making standard study designs for systems of nonlinear equations infeasible. Unfortunately, tools to optimize study design for systems of ODEs are not widely known and the feasibility of parameter estimation with chosen the chosen study design often goes unchecked. We review the impact experimental design choices make on parameter estimation with fixed resources and available tools to optimally choose study designs for a pre-specified system of equations and parameters of interest with two mindsets: “locally optimum” designs and Bayesian priors for parameters.