A0947
Title: Model selection for unit-root time series with many predictors
Authors: Ching-Kang Ing - National Tsing Hua University (Taiwan) [presenting]
Abstract: Model selection is studied for a general unit-root time series when many predictors are present. A new model selection algorithm called FHTD is proposed that leverages the advantages of forward stepwise regression (FSR), the high-dimensional information criterion (HDIC), a backward elimination method based on HDIC, and a data-driven thresholding (DDT) approach. By deriving a new functional central limit theorem for multivariate linear processes, along with a uniform lower bound for the minimum eigenvalue of the sample covariance matrices of the series under study, the sure screening property of FSR and the selection consistency of FHTD under some mild assumptions that allow for unknown locations and multiplicities of the characteristic roots on the unit circle and conditional heteroscedasticity in the predictors and errors, are established. The simulation results corroborate the theoretical properties and show the superior performance of FHTD in model selection. FHTD is also applied to U.S. monthly housing starts and unemployment data to demonstrate its usefulness in practice.
