A0691
Title: Regularized estimation of loading matrix in factor models for high-dimensional time series
Authors: Xialu Liu - San Diego State University (United States) [presenting]
Xin Wang - San Diego State University (United States)
Abstract: High-dimensional data analysis using traditional models suffers from over-parameterization. There are two types of techniques commonly used to reduce the number of parameters involved- (1) regularization and (2) dimension reduction. The purpose is to combine them by imposing a sparse factor structure and propose a regularized estimator to further reduce the number of parameters in factor models. One issue for factor models is that factors are hard to interpret since both factors and the loading matrix are unobserved. When estimating the loading matrix, a penalty term is introduced for a sparse estimate. Hence, each factor only drives a smaller subset of time series that exhibit the strongest correlation, making factor interpretation much clearer. The theoretical properties of the proposed estimator are investigated. The simulation results are presented to confirm that our algorithm performs well. The method is applied to Hawaii tourism data. Results show that two groups of people driving the number of domestic tourists in Hawaii are visitors from states in high latitudes (factor 1) and visitors from inland or states in low latitudes (factor 2). It reveals two main reasons that people visit Hawaii to (1) escape the cold and (2) enjoy the beach and water activities.
