Experimental Design in Validating Models of Infectious Diseases
David M. Bortz
Department of Applied Mathematics<
University of Colorado
In the mathematical modeling of infectious diseases, it is always
a challenge to construct a good model. In particular, we are
interested in developing models for early infection HIV
pathogenesis and Staphylococcus epiermidis bacteremia. We have
proposed a methodology for systematic model selection among a set
of candidate models which balances goodness-of-fit with a measure
of statistical sophistication (Bortz & Nelson, Bull. of Math.
Bio., 2006). Within the experimental design literature there are
criteria which can dictate a sampling strategy to minimize either
parameter estimate error or minimize model prediction error.
Intriguingly, these criteria are directly related to a measure of
the statistical complexity of the model. We will present results
illustrating selection between competing models for HIV
pathogenesis and Staphylococcus epiermidis bacteremia as well as
how to identify an optimal sampling strategy for validated
incorporation of salient immune system features into the model.
Back to the