比例优势模型实现ROC分析的方法及其应用前景分析

目的 探讨比例优势模型在ROC分析中的应用前景。方法 比较比例优势模型与双正态模型等经典方法所计算的ROC曲线下面积及其标准误；采用灵敏度残差平方和与决定系数两个指标评价参数模型的拟合优度。结果 在一般情况下，由比例优势模型所得到的Roe曲线指标结果与经典方法很接近；对于有序分类资料和连续型资料，该模型的拟合效果均较好；但由于该模型获得的Roe曲线形状单一，有些情况下该模型的拟合不理想。结论 与经典的方法相比，比例优势模型有其自身的特点，实际应用时应慎重做出选择。

The Proportional Odds Model and its Prospect Applied to ROC Analysis

Objective Explore the role of proportional odds model in ROC analysis. Methods Compare the area under ROC curve Of proportional odds model and its standard errors with of common methods, such as binormal model;evaluate the goodness of fit for the parametric models using the sum of square of residuals of sensitivity and determination coefficient between actual sensitivity and estimated sensitivity. Results The values of the measures of ROC analysis from the proportional odds model are similar with of common methods; whether ordinal category data or continuous data, the effect of fit for the model are generally good. Conclusion Comparing with common methods of ROC analysis, the proportional odds model have itself characteristic. It is an alternative method for ROC analysis in practice.