Work

Variable selection via fused sparse-group lasso penalized multi-state models incorporating molecular data

Year: 2024

Type: preprint

Abstract: In multi-state models based on high-dimensional data, effective modeling strategiesare required to determine an optimal, ideally parsimonious model. In particular, linkingcovariate effects across transitions is needed to conduct joint variable selection.A useful technique to reduce model complexity is to address homogeneous covariateeffects for distinct transitions. We integrate this approach to data-driven variableselection by extended regularization methods within multi-state model building.Wepropose the fused sparse-group lasso (FSGL) penalized Cox-type regression in theframework of multi-state models combining the penalization concepts of pairwisedifferences of covariate effects along with transition grouping. For optimization,we adapt the alternating direction method of multipliers (ADMM) algorithm totransition-specific hazards regression in the multi-state setting. In a simulation studyand application to acute myeloid leukemia (AML) data, we evaluate the algorithm’sability to select a sparse model incorporating relevant transition-specific effectsand similar cross-transition effects. We investigate settings in which the combinedpenalty is beneficial compared to global lasso regularization.

Source: arXiv (Cornell University)

DOI: 10.48550/arxiv.2411.17394

Authors: Miah, Kaya, Goeman, Jelle J., Putter, Hein, Kopp-Schneider, Annette, Benner, Axel

Institutions: German Cancer Research Center, Heidelberg University

Language: English

Topic: Statistical Methods and Inference

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

SDG: Good health and well-being

Open Access status: green

Funder: Deutsche Forschungsgemeinschaft