儿童智力低下危险因素研究中1:1配对Logistic回归的方法适应性检验

Testing on Applicability of Muti-factor Stepwise Logistic Regression

目的:探讨Logistic回归在多元逐步回归前方法适应条件,为该方法的正确应用提供借鉴。方法:选取山东潍坊市确诊的儿童智力低下患者和对照作1:1配对病例对照研究,运用条件Logistic回归的条件适应性检验和回归诊断进行分析判断。结果:logitp值与各变量之间线性关系判断显示除X24(孕前母亲吸烟指数)需要进行函数变换外,其余各变量均初步符合Logistic回归的应用条件;进行函数变换后,又进行了过离散、特异值、空单元、多元共线性等回归诊断,结果提示不存在数据结构问题,可以进入多元逐步Logistic回归。结论:从线性检验、函数变换和回归诊断角度进行多元逐步Logistic回归应用条件检验,以使多元逐步Logistic回归的结果和结论更严谨、更具有说服力。

Objective: To explore the applicability before muti-factor stepwise regression of Logistic Regression, which providing use for reference for proper application. Methods: Choosing cases and controls of child mental retardation as studying objective and analyzing on the applicability testing and regression diagnosis of Logistic Regression. Results: The linearity estimation between logitp value and variables shows that the other variables primarily accord with the applying conditions of Logistic Regression, except X24 needing function transform. After regression diagnosis of over-dispersion, outliers, zero cell count, multicollinearity, the results shows that there are no data structure problems. So muti-factor stepwise Logistic Regression can go on. Conclusions: The study tests applicability of muti-factor stepwise Logistic Regression by linearity testing, function transform and regression diagnosis, which makes the results and conclusions more precise and persuasive.