基于指数模型的R=P(Y<X<Z)统计推断

Statistical Inference for R=P(Y<X<Z)of Exponential Distribution

文章讨论了接收器工作特性曲线(ROC)指标R=P(Y<X<Z)的统计推断问题,分别给出了R的不同点估计和区间估计,其中点估计包括最大似然估计和贝叶斯估计,区间估计包括正态近似区间估计、精确区间估计以及贝叶斯可信度置信区间估计。通过数值模拟,比较了上述估计方法的优劣,并且给出了选择不同估计方法的建议,数值模拟结果表明上述估计方法是可行的。

In this article,the statistical inference of the receiver-operating characteristic(ROC)curves R=P(Y<X<Z)is provided.Different point estimations and interval estimations are proposed.The point estimators obtained include maximum likelihood and Bayesian estimators.The interval estimations obtained include approximate intervals,exact confidence intervals and Bayesian credible intervals.Different methods and the corresponding confidence intervals are compared using simulation studies,and suggestion is given to choose different estimation method.It shows that these methods are applicable.