Abstract: In this note we delineate conditions under which continuous time stochastic processes can be identified from discrete data. The identification problem is approached in a novel way. The distribution of the observed stochastic process is expressed as the underlying true distribution, f, transformed by some operator, T. Using a generalization of the Taylor series expansion, the transformed function T f can often be expressed as a linear combination of the original function f. By combining the information across a large number of such transformations, the original measurable function of interest can be recovered.

Full paper (92 KB PDF)