Department of Statistics
March 25, 2014
Stochastic inference of dynamic system models: from single-molecule experiments to statistical estimation
Dynamic systems, often described by coupled differential equations, are used in modeling diverse behaviors in a wide variety of scientific areas. In this talk we will consider their assessment and calibration in light of experimental/observational data. For the assessment, we explore how the deterministic dynamic system models reconcile with stochastic observations, using recent single-molecule experiments on enzymatic reactions as an example. For the calibration, we will propose a new inference method for the parameter estimation of dynamic systems. The new method employs Gaussian processes to mirror a dynamic system and offers large savings of computational time while still retains high estimation accuracy. Numerical examples will be used to illustrate our estimation method.