A regularized bridge sampler for sparsely sampled diffusions

摘要: Sparsely sampled diffusion processes, in this paper interpreted as data sparsely time relative to the constant, is a challenging statistical problem. Most approximations of transition kernel are derived under assumption that frequently and these often severely biased for data. Monte Carlo methods can be used problem density estimated with arbitrary accuracy regardless sampling frequency, but computationally expensive or even prohibited unless effective variance reduction applied. The state art technique processes Durham-Gallant (modified) bridge sampler. Their importance sampler using linearized, Gaussian approximation dynamics, have proved successful However, not valid data. We present flexible, alternative derivation modified multivariate, discretely observed models modify it by taking uncertainty into account. The resulting viewed combination basic sampler, most appropriate given problem, while still being efficient. Our providing