Speaker: Dr. Zhentao Shi (Chinese University of Hong Kong)
On LASSO for Predictive Regression
Room A619, Teaching A
Dr. Zhentao Shi (Chinese University of Hong Kong)
Dr. Shi is Assistant Professor at the Department of Economics. He specializes in econometric theory, in particular the estimation and inference in high-dimensional econometric models. He is also interested in empirical studies in finance, trade, labor and industrial organization.
This paper studies possibilities of using shrinkage methods for predictive regression. The variable selection in predictive regression is important since there is a variety of potential predictor variables. The commonly used predictors typically have various degrees of persistence, and exhibit low signal strength in explaining the dependent variable. We investigate the pitfalls and possibilities of the LASSO methods in this framework of predictive regression with mixed degrees of persistence. We show that the adaptive LASSO methods have the consistent variable selection and the oracle properties under the presence of stationary, unit root and cointegrated predictors. The conventional LASSO methods under this environment are also studied, signifying some practical concerns. Exploratory simulation results are reported, and some empirical practices are performed for illustration.