MDP for integral functionals of fast and slow processes with averaging (Q2485478)

The moderate deviation principle (MDP) is established for an integral functional of a stochastic system of ``Freidlin's type'', with fast and slow components. An advantage of this approach for computations is a much more easily computable rate function, in comparison to the ``standard'' large deviation (LD) scale. This provides, of course, certain asymptotics in different scales of parameters for MDP and LDP; however, e.g., in some statistical problems both scales may be suitable, so that the computational convenience can be, indeed, important. Recent advances in large deviations, due to A. Pukhalskij, and Poisson equations in unbounded Euclidean spaces are essential components of the method.