Title: On a data-driven semiparametric nonlinear model with penalized spatio-temporal lag interactions
Authors: Dawlah Alsulami - King Abdulaziz University (Saudi Arabia)
Zhenyu Jiang - University of Southampton (United Kingdom)
Zudi Lu - University of Southampton (United Kingdom) [presenting]
Jun Zhu - University of Wisconsin (United States)
Abstract: Studying a possibly nonlinear impact of consumer price index (CPI) on the housing price at a state level in the USA, ignoring or misspecifying the temporal lag effects of the housing price both from the own state and the neighboring states, will result in a biased modelling. We therefore propose a data-driven semiparametric nonlinear time series regression model that accounts for spatio-temporal lag interactions in both space and time. A semiparametric penalised estimation procedure is suggested by utilising adaptive lasso to estimate the most important spatio-temporal lag interactions. Theoretical justification for the estimation procedure is developed. Empirical application to a USA data set demonstrates that the proposed method can substantially improve the estimation of the nonlinear impact of the monthly increment of CPI on the housing price return at a state level, with interesting spatio-temporal lag interactions identified.