springer

springer implements bi-level variable selection using sparse group penalization to analyze high-dimensional longitudinal data for gene-environment interaction studies.

Key Features:

• Bi-Level Variable Selection: Performs simultaneous penalization at both group and individual levels using sparse group variable selection to handle hierarchical variable structures.
• Quadratic Inference Function (QIF) Framework: Implements a QIF-based penalization method (Zhou et al. 2022) to accommodate longitudinal responses and time-dependent data.
• Generalized Estimating Equation (GEE)-Based Penalization: Provides a GEE-based sparse group penalization approach to handle correlated observations in longitudinal studies.
• Alternative Methods: Offers methods that apply exclusively group-level or exclusively individual-level penalization.
• Response Types: Applicable to continuous, binary, and survival responses in high-dimensional settings.

Scientific Applications:

• Gene-Environment Interaction Studies: Identifies genetic and environmental factors and their interactions over time in longitudinal datasets.
• Genomic High-Dimensional Analysis: Analyzes genomic studies with many predictors and grouped variable structures.
• Longitudinal Modeling of Time-Dependent Outcomes: Models continuous, binary, or survival outcomes with correlated repeated measures.

Methodology:

Implements sparse group penalization and bi-level variable selection within QIF and GEE frameworks for longitudinal responses, reduces model complexity by selecting relevant variables, and incorporates the QIF-based penalization approach of Zhou et al. (2022).