Dimension Reduction Methods for Multivariate Time Series
by wbnicholson for R Project for Statistical Computing
Multivariate time series are ubiquituous within macroeconomic forecasting. The vector autoregression, the canonical modeling approach, is heavily overparameterized and is intractible in high dimensions. Our project aims to create an easily accessible R package which allows for the estimation of high-dimensional vector autoregressions by incorporating dimension reduction methods from the statistical regularization literature into a multivariate time series setting.