Two Sample Comparison Based on Empirical Likelihood Ratio Test for Right Censored Data

• Main Authors: HUNG, YU-JING, 洪于景
• Other Authors: HSIEH, JIN-JIAN
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/49wg2c

碩士 === 國立中正大學 === 數學系統計科學研究所 === 105 === It has been a common question for detecting two-sample survival curves in biomedical studies and clinical field. Among those proposed tests, the log rank test is the most popular approach. However, it fails to perform well under conditions of crossing survival functions. Moreover, there is no approach in the possession of preponderance for all situations in evidence. Therefore, we utilize the decent properties of empirical likelihood ratio to develop other methods. We also conduct several types of crossing survival curves and conduct the simulations to investigate the power and type I error rate. Then we compare the proposed approaches with some existing methods such as log rank test, Gehan-Wilcoxon test and Tarone-Ware test, etc. From simulations, we recommend EL-based method for two survival curves with two crossing points. The EL-based approaches are also possessed of adaptable type I error. Eventually, we analyze two medical data examples for illustrations.