A local vector autoregressive framework and its applications to multivariate time series monitoring and forecasting

Abstract

Our proposed local vector autoregressive (LVAR) model has time-varying parameters that allow it to be safely used in both stationary and non-stationary situations. The estimation is conducted over an interval of local homogeneity where the parameters are approximately constant. The local interval is identified in a sequential testing procedure. Numerical analysis and real data applications are conducted to illustrate the monitoring function and forecast performance of the proposed model.