No label defined (Q4538693)

A computational method of conditional moments of \(d\)-dimensional diffusion processes is considered. The author shows that conditional moments of a diffusion process can be approximated by the solution to an ordinary differential equation whose coefficients are characterized by coefficients of the stochastic differential equation of the investigated process. It is shown that the method gives the exact conditional moments if a stochastic differential equation has a special form. The numerical experiment of estimating the parameters of a nonlinear stochastic differential equation from discrete observation shows that the maximum likelihood estimation with the computed conditional moments performs better than a conventional discretization approach.