对COX模型多因子共线性处理方法的改进

Improvement of COXRegression Modelfor Handling Data with Multifactors Collinearity

目的：探讨处理ＣＯＸ 模型多因子共线性的改进方法。方法：与逐步回归、主成分分析比较，采用因子分析法将多因子分解为公因子和特殊因子两部分，对这两部分信息分别建立ＣＯＸ模型。结论：可以消除由于共线性的存在而造成应用逐步回归时产生的矛盾，避免应用主成分分析时造成信息的损失，使模型的解释更加全面。

Objective:Toimprove the Cox regression modelfor handling data with multifactors collinearity. Methods: With factoranalysis procedure,multifactors were di videdinto com mon factorsand uniquefactors.The Cox regression models werethen es tablished based on thecom monfactorsand uniquefactorsrespectivelyand compared with stepwiseregression and principalcomponentanalysis method .Result:Theimproved Cox regression modelcouldresolvethecontradictory produced byapplyingstepwiseregressionanalysisand avoid losing information when using principal component analysis. There fore,the explanation fortheeffects ofeach factor was more completed.