A Perturbation Subsampling Method for Massive Censored Data

Area of Science:

• Statistics
• Biostatistics
• Computational Biology

Background:

• Large-scale data analysis is increasingly prevalent due to advancements in information technology.
• Statistical analysis of large datasets often requires computationally intensive subsampling methods.
• Calculating sampling probabilities for each observation in subsampling can be a significant computational bottleneck.

Purpose of the Study:

• To extend the perturbed subsampling approach to the Cox proportional hazards model.
• To develop an efficient perturbed subsampling algorithm for analyzing large-scale survival data.
• To address the computational challenges associated with survival analysis on big data.

Main Methods:

• Extension of the perturbed subsampling technique.
• Development of a novel perturbed subsampling algorithm tailored for the Cox proportional hazards model.
• Evaluation through simulation studies and real-data analysis.

Main Results:

• The proposed perturbed subsampling algorithm effectively handles large-scale survival data.
• The method significantly reduces computational intensity compared to traditional subsampling approaches.
• Simulation studies and real-data analysis confirm the method's effectiveness.

Conclusions:

• The perturbed subsampling approach provides an efficient solution for survival analysis with large-scale data.
• This algorithm enhances the feasibility of applying the Cox proportional hazards model to big data.
• The developed method offers a valuable tool for biostatisticians and data scientists working with extensive survival datasets.