svreg

svreg implements structural varying-coefficient regression for model selection in varying-coefficient models with structured main predictors or modifying variables, enabling identification of how relationships between predictors and outcomes vary across groups or conditions.

Key Features:

• Structural Varying-Coefficient Regression: Implements varying-coefficient models that allow predictor-outcome relationships to change across groups or conditions.
• Model Selection with Structured Variables: Performs model selection while accounting for pre-specified group structures among predictors or modifying variables.
• Pliable Lasso (plasso): Includes pliable lasso (plasso), an extension of lasso that enables variable selection across different groups or conditions.
• Novel Variable Selection Method: Employs a variable selection approach that consistently identifies relevant variables, screens out irrelevant ones, and improves sensitivity, false discovery rates, and prediction accuracy.

Scientific Applications:

• Neurological Research: Applied to Huntington disease studies to identify brain regions associated with motor impairment and to assess how these associations vary by patient group.
• Personalized Medicine: Enables identification of interaction effects between disease severity and biological markers (e.g., brain regions) to support development of customized treatment plans.

Methodology:

Implements structural varying-coefficient regression and model selection with structured main predictors or modifying variables, incorporates pliable lasso (plasso), and applies a novel variable selection procedure that was empirically validated to improve sensitivity, reduce false discovery rates, and enhance prediction accuracy, including detection of interaction effects between disease severity and brain regions.