Vector Autoregression: Modeling Vector-Valued Time Series

This lecture by the instructor covers the concept of Vector Autoregression (VAR) for modeling vector-valued time series. Starting from the basics of AR processes, the lecture explains how to extend them to vector-valued processes using the VAR(p) model. The stability of VAR(p) models is discussed, along with examples and calculations of reverse characteristic polynomials. The lecture also delves into the computation of auto-covariance matrices, Yule-Walker equations, and autocorrelations for VAR processes. The presentation concludes with a summary of the key topics covered, including linear processes, spectral decomposition, estimation methods, forecasting, and multivariate models.