Related Experiment Videos
An Exploratory Stability Selection (ESS) framework for robust predictor discovery: An application to physical
Marissa C Ashner1, Virginia B Kraus2, Heather E Whitson3
1Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, NC, USA; Center for the Study of Aging and Human Development, Duke University School of Medicine, Durham, NC, USA.
Experimental Gerontology
|August 10, 2026
Summary
A new statistical framework, Exploratory Stability Selection (ESS), helps identify reliable predictors of physical resilience. This method enhances precision in aging and recovery research by stabilizing variable selection in complex datasets.
Area of Science:
- Gerontology and Rehabilitation Science
- Biostatistics and Data Science
- Biomarker Discovery
Background:
- Identifying predictors of physical resilience is crucial for personalized aging and recovery strategies.
- High-dimensional data presents analytical challenges like correlation and instability, complicating variable selection.
- Robust and transparent methods are needed to discover reliable biological and clinical signals.
Purpose of the Study:
- To introduce the Exploratory Stability Selection (ESS) framework for hypothesis-generating predictor discovery in complex datasets.
- To demonstrate ESS's utility in identifying stable and context-dependent predictors of physical resilience.
- To provide a method for prioritizing and validating candidate predictors in aging and recovery research.
Main Methods:
- Developed a resampling-based statistical framework (ESS) integrating multiple resampling strategies and sparsity levels.
- Applied ESS to clinical, plasma biomarker, and combined predictor sets from the PRIME-KNEE study.
- Evaluated predictor stability and competitiveness for resilient recovery trajectories in older adults undergoing knee arthroplasty.
Main Results:
- ESS successfully distinguished stable predictors from context-dependent signals influenced by predictor competition and analytic choices.
- The framework demonstrated robustness across different data perturbations and analytic configurations.
- Analysis of PRIME-KNEE data highlighted stable predictors associated with resilient recovery from pain interference.
Conclusions:
- The ESS framework offers a transparent and stable approach for variable selection in complex, high-dimensional data.
- ESS is well-suited for hypothesis generation in resilience research and other aging-related applications.
- This method aids in prioritizing and validating candidate predictors for advancing precision medicine in recovery and aging.