Estimation of stochastic models under indirect observability
Arjun Beri (Mathematical Biosciences Institute, The Ohio State University)
(September 22, 2011 10:30 AM - 11:30 AM)
I will present the problem of adequate data subsampling for asymptotically consistent parametric estimation of unobservable stochastic differential equations (SDEs) when data are generated by multiscale dynamic systems approximated by these SDEs. The challenge is that the approximation accuracy is scale dependent, and degrades at very small scales. Data from multiscale dynamics systems, namely the Additive Triad model, will be used to illustrate this subsampling problem. I will also indicate the general framework for estimation under this indirect observability and present practical numerical techniques to identify the correct subsampling regime to construct bias-corrected estimators.