Cox风险比例回归的惩罚函数变量选择

Selection of Penalty Function Variables Based on Cox Risk Proportional Regression

本文主要通过比较Cox风险比例生存回归模型中不同的惩罚函数变量选择方法,应用惩罚函数进行变量选择和参数估计,以Van′t Veer(2002)等学者的乳腺癌数据集为例,在Cox风险比例回归模型中将基于惩罚函数各种变量选择方法应用于肿瘤遗传学数据分析中。研究表明,Cox风险比例回归模型中基于双层变量选择cMCP具有有效的降维能力--只要删失比例不高,均能达到优良的效果。

This article mainly compares different penalty function variable selection methods in the Cox risk proportional survival regression model through simulated data.Penalty function for variable selection and parameter estimation is used to study the breast cancer data set of Van't Veer(2002)as an example.In the Cox risk proportional regression model,various variable selection methods based on the penalty function are applied to the analysis of tumor genetics data.Studies results show that the choice of cMCP based on two-level variables in the Cox risk proportional regression model has an effective dimensionality reduction capability.It can achieve good results as long as the censoring ratio is not high.