STATISTICS SEMINAR
Wednesday, May 9th, 2012 at
10:00am, MSB 1147 (Colloquium Room)
Refreshments prior to seminar in MSB 4110 (Statistics Lounge)
Speaker: Edward
Ionides (University of
Michigan)
Title: Inference for
partially observed stochastic dynamic systems
Abstract: Inferential challenges arise in the
study of biological dynamic systems due to the combination of stochasticity,
nonlinearity, measurement error, unobserved variables, unknown system
parameters, and unknown system mechanisms. I will discuss statistical
methodology developed to address these challenges, with particular reference to
pathogen/host systems (i.e., disease transmission). I will focus on methodology
which is based on simulations from a numerical model; such methodology is said
to have the plug-and-play property. Plug-and-play approaches free the modeler
from an obligation to work with models for which transition probabilities are
analytically tractable. A recent advance in plug-and-play likelihood-based
inference for general partially observed Markov process models has been
provided by the iterated filtering algorithm. I will discuss the theory and
practice of iterated filtering.
