Workshop on Statistical Methods for Dynamic
Vancouver, June 4-6 2009
System Models
Progress in scientific investigation is accelerated by the ability to formulate and compare multiple mechanistic models. Most existing methods for statistical inference on mechanistic models place rather severe restrictions on the form of the models that can be entertained. "Plug-and-play" methods, by contrast, require only that one be able to simulate a stochastic dynamical system and are thus free of such restrictions. I will point out several of these methods and describe one---Iterated Filtering---in some detail, using examples to show that one can use it to ask and answer questions previously unaddressable. Along the way, I will introduce an R package, pomp, which provides a general modeling and inference framework for stochastic dynamical systems.
Plug-and-play inference for stochastic dynamical systems
Assistant Professor of Ecology & Evolutionary Biology and Mathematics
University of Michigan
Ann Arbor, Michigan
