Department of Biostatistics Seminar/Workshop Series
Biostatistics Student Research Forum:
Optimal design for nonlinear models
Ryan T. Jarrett, Ph.D. Candidate, Department of Biostatistics
Vanderbilt University, School of Medicine
An optimal design will gain more information about parameters of interest than a suboptimal design with the same number of observations, or equivalently, it will accumulate the same amount of information with fewer observations. By designing a study optimally, we can therefore achieve the same results faster, with less cost, or with fewer patients than we could with a suboptimally-designed study. We will review some of the work done in optimal design that seeks to answer the question “Given a pre-specified model and estimator(s) of interest, where should we take measurements so as to best estimate parameters of interest with the most precision?” We will discuss ways to quantify and compare information measured across differing designs, how to identify the support points that hold the most information, and complications associated with designs for nonlinear models. We conclude by applying this theory to an example in pharmacokinetics, in which we are interested in identifying times to draw blood in order to measure the time-concentration curve for an individual patient.