A Time-Lagged Penalized Regression Model and Applications to Economic Modeling

• Published in: Journal of statistical theory and practice Vol. 18; no. 1
• Main Authors: Sun, Mingwei, Zheng, Rong
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.03.2024
• Subjects: Mathematics and Statistics; Original Article; Probability Theory and Stochastic Processes; Statistical Theory and Methods; Statistics; High-dimensional data; Time-delayed effect; Penalized regression; Economic modeling; Time lag; Variable selection
• ISSN: 1559-8608; 1559-8616
• DOI: 10.1007/s42519-023-00354-3

In the arena of high-dimensional data analysis, variable selection has emerged as a significant subject. The simultaneous accomplishment of variable selection and coefficient estimation has been realized through penalized regression, which has seen varied degrees of success across numerous fields in recent years. This paper introduces a time-lagged penalized regression model that integrates the advantages of penalized regressions and accommodates time-delayed effects present within a dataset. The model discerns the lag times that maximize the correlations between the dependent and each independent variable, subsequently implementing a transformation of the data based on these lags. The model’s application is demonstrated in the context of economic modeling, a field often characterized by an abundance of lagged variables. The ensuing results reveal the model’s capacity to uncover hidden variables, overlooked by linear penalized regression models, and to exhibit superior predictive performance. The proposed model thus offers a novel method for a more comprehensive and objective elucidation of data.