为了进一步提高pAUC(Partial area under curve)估计精度和医学诊断测试精确性,提出了一种基于密度比模型的pAUC半参数估计方法,并从理论和仿真两个方面研究其性质.首先,根据密度比模型,用半参数极大似然估计方法得到了 pAUC半参数估计量,并用大样本理论分析了它的统计性能;然后,对pAUC半参数估计方法在实际应用中的性能进行了仿真,并与现有精度较高的pAUC非参数估计方法进行比较.研究发现,pAUC半参数估计量不仅具有相合渐近正态性等重要的统计性质,而且比已有的非参数pAUC估计量具有更高的渐近估计效率和精确度.将该pAUC半参数估计方法应用于乳腺癌诊断模型的筛选,得到了一个预测精度更高的新乳腺癌诊断模型,结果表明该方法在实际应用中能提高医学诊断测试的精度.
A semi-parametric estimation method for pAUC based on the density ratio model and its application
In order to further improve the estimation accuracy of pAUC(partial area under curve)and the accuracy of medical diagnosis tests,a semi-parameter estimation method of pAUC based on density ratio model is proposed,and its properties are studied from both theoretical and simulation aspects.Firstly,according to the density ratio model,the semi-parametric maximum likelihood estimator of pAUC is obtained based on the semi-parametric maximum likelihood estimation method,and its statistical performance is analyzed by using the large sample theory.Then,the performance of the pAUC semi-parametric estimation method in practical application is simulated and compared with the existing non-parametric estimation method in term of accuracy.It is found that not only the semi-parametric pAUC estimator has important statistical properties such as consistent asymptotic normality,but also it has higher asymptotic estimation efficiency and accuracy than the existing nonparametric pAUC estimator.The semi-parameter estimation method for pAUC is applied to the screening of breast cancer diagnosis models,and a new breast cancer diagnosis model with higher prediction accuracy is obtained.The result shows that the proposed method can improve the accuracy of medical diagnosis tests in practical applications.
pAUCsemi-parametric estimatiordensity ratio modelasymptotic normalitymedical diagnosis
NETL
NSTL
万方数据
