论文部分内容阅读
目的:介绍净重新分类指数与整体鉴别指数,阐明两个统计指标在医学研究中普遍使用的Logistic回归模型预测效果评价中的意义与实现.方法:给出净重新分类指数与整体鉴别指数的定义式,结合实例说明在R中的计算实现,并与常用的预测效果评价指标:受试者工作特征曲线,曲线下面积,以及灵敏度、特异度等进行对比,阐述其应用方面的特点.结果:在Logistic回归模型应用中,净重新分类指数与整体鉴别指数可以量化出模型筛选中新模型的改进效果,在R软件中可以方便地获得此效果的估计值及假设检验结果.结论:净重新分类指数与整体鉴别指数定量地反映了模型优化前后的差异,具有实际运用价值.运用R软件可完成相关计算.“,”Objective:To introduce net reclassification improvement and integrated discrimination improvement, and clarify the significance and application of these two statistical indexes in the evaluation of prediction effect of logistic regression models commonly used in medical research. Methods:Definitions of net reclassification improvement and integrated discrimination im-provement were given with examples to illustrate the computation of the two indexes in R. We compared them with commonly used indexes comparing prediction effect such as receiver operating characteristic curve, area under the curve, sensitivity and specificity, so as to elaborate their characteristics in application. Results:In logistic regression models, net reclassification im-provement and integrated discrimination improvement could quantify the improvement of prediction effect from old to new mod-els. Estimated values and hypothesis testing results of this improvement can be easily obtained in R software. Conclusion:Net reclassification improvement and integrated discrimination improvement quantitatively reflect the difference before and after the model optimization, and have practical value. R software can be used to complete the calculations.