Copula Models with Filtered Time Series

• Activity Number: 626
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Zhijie Xiao*+
• Companies: Boston College
• Keywords: Copula ; Time Series ; Semiparametric ; Filter ; Nostationarity

Abstract

Copula models are widely used to capture nonlinear dynamics in time series. Applications of copula models depend on the assumption of stationarity. An important issue in practice is that nonstationarity and nonlinearity may occur simultaneously. In this paper, we study copula based time series that are potentially nonstationary. We consider three-step estimation procedures where nonstationarity is removed from a first stage filtration, and then two-step copula-based analysis is applied to the filtered data. Estimation and inference of the copula models for filtered time series are studied. Both parametric models, where the marginal distributions belong to some parametric families, and the semiparametric models where the marginal distributions are nopnparametrically modelled, are investigated. We show that the limiting distribution of the semiparametric estimator is not affected by the preliminary filtration procedure. Such a property does not hold for parametric models in general. This suggests using the more robust semiparametric method in practice.