Marginal false discovery rate for a penalized transformation survival model

Area of Science:

• Biostatistics
• Statistical Inference
• High-Dimensional Data Analysis

Background:

• Survival analysis with many variables is increasingly common.
• Existing methods focus on estimation and variable selection using regularization.
• Limited inference methods exist for high-dimensional survival data, often tied to specific models like Cox regression.

Purpose of the Study:

• To develop a flexible and statistically valid inference approach for survival analysis with moderate/high dimensional covariates.
• To extend inference capabilities beyond traditional Cox models.
• To provide a robust alternative for practical data analysis.

Main Methods:

• Utilized transformation models, known for their robustness and flexibility.
• Expanded the scope of inference beyond specific parametric models.
• Established rigorous statistical validity for the proposed methods.

Main Results:

• Developed a new inference approach for regularized survival data.
• Demonstrated the applicability and robustness of transformation models for inference.
• Successfully applied the methods in two real-world data analyses.

Conclusions:

• An alternative inference approach for high-dimensional survival analysis has been established.
• The transformation model offers a more flexible framework for statistical inference in this context.
• The developed methods provide statistically sound and practically useful tools for analyzing complex survival data.