重複事件資料常見於長期追蹤研究,在醫學領域的實例如:HIV患者重複遭受伺機性感染的記錄、膀胱癌患者的術後復發記錄。重複事件過程可能因發生終止事件,如:死亡,而終止後續的復發事件;估計累積復發率經常是分析重複事件資料的主要目的之一,必須在重複事件過程與終止事件時間獨立的條件成立下方可得到累積復發率的無母數估計量,如:Nelson-Aalen估計量。但當重複事件過程與終止事件時間相依,Nelson-Aalen估計量是有偏的。本文推廣過去文獻以條件Kendall’s tau所建立的 統計量,發展無母數方法檢定重複事件過程與終止事件過程的獨立性,並推導檢定統計量在虛無假說下的大樣本性質。在模擬試驗中分析所提出之檢定統計量於不同情境下的表現。最後,以一組實際資料示範檢定方法的應用。
Recurrent event data are often encountered in longitudinal follow-up studies. Examples in medical research include the repetitions of opportunistic infections in HIV-infected patients, or the repeated tumor resections of patients with postoperative superficial bladder tumor. In many cases, there exists a terminal event (e.g., death), which precludes the subsequent recurrent events. Estimating the cumulative recurrent rate function is often one of the aims of analying recurrent event data. The cumulative recurrent rate function is nonparametrically identifiable if the independence of recurrent event process and terminal event time holds. When the recurrent event process and terminal event time is dependent the nonparametric estimate of cumulative recurrent rate function, such as Nelson-Aalen estimate, is biased. We propose nonparametric tests for independence of recurrent event process and terminal event time by extending the -statistic developed from a conditional Kendall’s tau. The asymptotic distributions of the test statistics are derived under the null. The performance of the proposed tests is evaluated by simulation studies under various scenarios. The application of the proposed method is illustrated with a real example.