Title: Mixed time aggregation of dynamic multivariate linear processes
Authors: Michael Thornton - University of York (United Kingdom) [presenting]
Abstract: The time aggregation of vector linear processes: (i) containing mixed stock-flow data; and, (ii) aggregated at mixed frequencies is explored, showing how the parameters of the underlying model translate into those of the equivalent model of the aggregate. Based on manipulations of a general state-space form, the results may be applied to a wide range of linear ARMAX processes, including the discrete representation of a continuous time process, and may be iterated to model multiple frequencies or aggregation schemes. Estimation via the Kalman-Bucy filter and via the ARMAX representation of the observable data is discussed.