logistic回归模型在ROC分析中的应用  

Application of logistic Model in ROC Curve Analysis

作　　者：陈卫中[1] 潘晓平[2] 倪宗瓒[2] 

机构地区：[1]成都医学院公共卫生学教研室,610081 [2]四川大学华西公共卫生学院卫生统计学教研室

出　　处：《中国卫生统计》2007年第1期22-24,共3页Chinese Journal of Health Statistics

摘　　要：目的探讨logistic回归模型在有协变量或多指标联合诊断试验ROC分析中的应用。方法根据疾病状态建立logistic回归模型,通过形成的预测概率或联合预测因子为分析指标,并结合非参数模型和双正态模型建立ROC曲线。结果通过实例阐述了整个分析过程,并说明了该试剂盒的有效性,同时利用两种模型得到了一致的结果。结论ROC分析中结合logistic回归模型简单有效,尤其适用于有协变量或多指标联合诊断试验的分析评价。To explore the application of logistic model in ROC curve analysis. Methods Based on the logistic model, combining predictors or probabilities were gained and applied in establishing empirical and binormal model ROC curves. Results An example for prediction of coronary heart disease was presented to illustrate the whole analysis steps. At the same time, the method showed conformably that the OxLDL was suitable for coronary heart disease screening. Conclusion Roe analyis based on the logistic model is easy and convenient, especially using in the screening test with covariates or multiple markers for classification.

关 键 词：诊断试验 ROC曲线 冠心病 logistic模 型双正态模型 

分 类 号：R195[医药卫生—卫生统计学]