CRoNoS & MDA 2019: Start Registration
View Submission - CRONOSMDA2019
A0150
Title: High frequency linear time series models and mixed frequency data Authors:  Manfred Deistler - Vienna University of Technology (Austria) [presenting]
Abstract: The focus is on the identification of multivariate linear dynamic models from so called mixed frequency (MF) data, i.e. from data where the univariate components of the time series are sampled at different frequencies; in economic applications this occurs if e.g. unemployment data are sampled monthly and GNP is available only quarterly. The interest is in the underlying ``high frequency'' (HF) model, i.e. in the model generating outputs at the highest sampling frequency. The model classes considered are multivariate AR and ARMA models (both with nonsingular and singular innovation variance) and linear dynamic factor models. We discuss problems of parameter identifiability and of estimation. In estimation in particular MLEs and EM algorithms are analyzed, both w.r.t their asymptotic and finite sample properties. The information loss due to MF data relative to HF data is discussed.