P. González Barquero, R. E. Lillo Rodríguez, Á. Méndez Civieta
Regularization techniques are essential in high-dimensional scenarios. With the rise of big data, these contexts are becoming more common and, although having large volumes of data is beneficial in terms of information, it poses multiple challenges from a statistical point of view.
This study focuses on the use of variable selection techniques in the high-dimensional context, specifically for Cox regression, where the problem is unfeasible since, without regularization, there is an infinite number of possible solutions for the regression coefficients. A variety of variable selection methods are studied and tested, including the proposal of an adaptive version of the Group linear algorithm with sparse principal decomposition (GLASP), which is an extension of the Sparse Group Lasso that computes groups automatically.
Additionally, an application to genomic data is included with the aim of determining which variables influence the survival of patients with Triple-negative breast cancer.
Keywords: Survival analysis, Cox regression, high-dimensional data
Scheduled
Big Data processing and analysis (TABiDa2)
June 10, 2025 5:10 PM
Sala 3. Maria Rúbies Garrofé