A1677
Title: Adaptive group fused lasso for panel threshold model
Authors: Lulu Wang - Beijing Normal-Hong Kong Baptist University (China) [presenting]
Zudi Lu - City University of Hong Kong (China)
Abstract: Panel threshold regression has been one of the most popular methods in nonlinear panel time series analysis. The most common method to determine the number of threshold parameters is using a bootstrap procedure to approximate the sampling distribution under the assumption of cross-sectional independence, which may work poorly when dealing with strong dependence in most climate and finance data. A panel threshold model is considered where both regressors and residuals are allowed to be cross-sectionally dependent via adaptive group fused Lasso. It is shown that with probability approaching one, the proposed method can correctly determine the unknown number of threshold parameters and estimate regression parameters consistently. Asymptotic theories of the Lasso estimators of regression coefficients are established. Simulation studies demonstrate that the proposed estimation method works well in finite samples under cross-sectionally dependent conditions.
