On Simulated Likelihood of Discretely Observed Diffusion Processes and Comparison to Closed-Form Approximation

摘要: This article focuses on two methods to approximate the log-likelihood of discretely observed univariate diffusions: (1) simulation approach using a modified Brownian bridge as importance sampler, and (2) closed-form approximation approach. For case constant volatility, we give theoretical justification sampler by showing that it is exactly bridge. We also discuss computational issues in such accelerating numerical variance stabilizing transformation, computing derivatives simulated log-likelihood, choosing initial values parameter estimates. The approaches are compared context financial applications under benchmark model which has an unknown transition density no analytical transformation. approximation, particularly second-order closed-form, found be computationally efficient very accurate when observation frequency ...