A1164
Title: Consistent variable selection for high-dimensional complex regression structures: A two-stage screening approach
Authors: Wei-Cheng Hsiao - Soochow University (Taiwan) [presenting]
Abstract: The focus is on the challenge of variable selection in high-dimensional settings where the data generating process exhibits complex regression structures beyond standard additive assumptions. In many real-world applications such as bioinformatics or environmental monitoring predictors often influence not only the central tendency but also the variability and joint dependency patterns of the response variable. To capture these features, a consistent model selection procedure, OHT, and its extension MOHT is proposed. This framework integrates the orthogonal greedy algorithm for sequential screening with the high-dimensional information criterion to optimize the model path, followed by a trimming step to ensure parsimony. The method is designed to effectively identify significant variables governing both non-linear effects and distributional properties. Simulation results demonstrate that the proposed algorithm significantly outperforms existing methods, such as iFORT and RAMP, in terms of variable coverage and mean squared prediction error.
