Realized volatility and parametric estimation of Heston SDEs

We present a detailed analysis of the Heston model with particular emphasis on the indirect observability framework for parameter estimation in the volatility equation. Since the volatility process is not directly observable, values of parameters have to be inferred from an approximating realized volatility process. In this paper we analytically establish criteria for the optimal sub-sampling of the realized volatility process depending on the size of the averaging window. Our analytical results are supplemented by extensive numerical investigation of the Heston model. Thus, our analytical and numerical results in this paper provide practical guidelines for selecting the frequency of estimation, size of the observational samples, etc. to achieve near-optimal parameter estimation accuracy.