目的 介绍广义相加模型(GAM)识别非线性相关及其在医学统计建模中的应用.方法 应用SAS软件PROC GAM模块识别实例数据结局变量与自变量之间的非线性相关,通过比较考虑该非线性相关和不考虑该非线性相关时多元线性回归和lo/istic回归模型的拟合和预测效果,阐明GAM识别非线性相关在统计建模中的重要性.结果 与不考虑非线性相关的模型相比,考虑非线性相关的模型拟合和预测效果更优.结论 合理使用GAM,在模型中纳入非线性成分,可改善回归模型的建模效果和预测精度.%Objective To introduce Generalized Additive Models(GAM) in identifying non-linear correlations and its application in statistical modeling for medical research data. Methods A dataset was used for modeling with SAS PROC GAM. Goodness of fit and prediction precision were compared between models with and without non-linear components. Results A non-linear correlation could be identified by GAM. Compared with models without non-linear components, goodness of fit and prediction precision were improved by involving non-linear components. Conclusion Models with non-linear components reflect a true relation-ship between dependent and independent variables and hence improve the predictive ability.
展开▼