WorksMathematics
Limit theorems and bias correction for the initial-noise-adjusted SIML estimator
Abstract
We propose a sine-based modification of the separating information maximum likelihood (SIML) estimator for integrated volatility and covariance under equidistant sampling with microstructure noise. The estimator allows for nonzero initial observation noise. We establish consistency under moment assumptions and asymptotic normality under first-order Malliavin regularity of the diffusion coefficients. We also construct a feasible noise-bias correction using an estimator of the unknown noise covariance. This correction permits wider spectral bandwidths for consistency and stable asymptotic normality.