A0173
Title: Semiparametrically efficient rank-based estimation for dynamic location and scale models
Authors: Davide La Vecchia - University of Geneva (Switzerland) [presenting]
Marc Hallin - Universite Libre de Bruxelles (Belgium)
Abstract: New rank-based procedures are introduced to conduct semiparametric inference on time series models. In the considered setting, the conditional location and scale of the process depend on an Euclidean parameter, while the innovation density is an infinite dimensional nuisance parameter. Easy-to-implement rank-based estimators (R-estimators) are derived and their properties are discussed, with emphasis on semiparametric efficiency and root-$n$ consistency even in the presence of misspecification. The developed methodology has a wide range of applications, including linear and nonlinear models, in either discrete- or continuous-time, with either homo- or heteroskedasticity. Numerical examples about the modeling of the two scale realized volatility illustrate the performances of the proposed R-estimators. Finally, some extensions related to constrained inference on conditional duration models for market microstructure analysis and multivariate time series are briefly discussed.
