清华大学、交通大学 统 计 学 研 究 所 专 题 演 讲 题 目: Exact Non-inferiority Tests for the Difference in Partial Paired Areas under the ROC Curves 主讲人: 刘仁沛博士 (台湾大学农艺学系与国卫院生统组合聘) 时 间: 96年6月1日（星期五）10：40 - 11：30 （上午10：20-10：40茶会於统计所821室举行） 地 点: 清大综合三馆837室 Abstract When the standard procedure is available for diagnosis of a certain disease, it is more reasonable and feasible to verify the diagnostic accuracy of the new test procedure is not inferior to that of the current standard procedure. The difference in paired areas under the receiver operating characteristic (ROC) curves is one of the most commonly employed measures for comparison of overall diagnostic accuracy. However, the partial area under the ROC curves can be a better and more informative summary index for some diagnostic procedures which require the false positive rate not exceeding a certain level. Nonparametric and maximum likelihood methods can be used to construct the non-inferiority tests based on difference in paired partial areas under ROC curves. However, their performance on size and power has not been thoroughly investigated. We propose to use the concept of generalized pivotal quantities to construct an exact non-inferiority test for the diagnostic accuracy based on the difference in partial paired areas under ROC curves. A simulation study is conducted to empirically investigate the size and power of the three methods for various combinations of sample sizes, values of the partial areas under the ROC curve and correlations. Simulation results demonstrate that the exact non-inferiority test based on the concept of generalized pivotal quantities not only adequately controls the size at the nominal level but also is uniformly more powerful than the other two methods. Numerical examples using published data illustrate the proposed method. 敬请公布 欢迎参加 --

※ 发信站: 批踢踢实业坊(ptt.cc) ◆ From: 140.113.114.165