Two-sample quantile residual life test with clustered survival data

• Main Authors: I-WENLAN, 藍翊文
• Other Authors: Yun-Chan Chi
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/thr8w9

碩士 === 國立成功大學 === 統計學系 === 105 === To make inference about the difference between two quantile residual lifetimes for clustered survival data, this thesis first extends two minimum dispersion test statistics to clustered survival data. However, the simulation results in Jeong et al. (2008) and Tsai et al. (2016) have shown that the type I error rates of these two minimum dispersion test statistics are too conservative for independent survival data. Therefore, this thesis proposes three Wald type test statistics based on the difference between two sample quantile residual lifetimes with different length-based variance estimators, which are proposed by Tsai et al. (2016). A simulation study is conducted to examine the type I error rates and power of the test statistics proposed in this thesis for clustered survival data. From the results of simulation study, the type I error rates of the three Wald type test statistics are close to the nominal type I error rate. Whereas, the type I error rates of the two minimum dispersion test statistics are very conservative. Furthermore, we compare the power of the five test statistics under Weibull distribution, and the result shows that the three Wald type test statistics are more powerful than the two minimum dispersion test statistics. Therefore, Wald type test statistics are recommended to be applied in practice when survival times follow Weibull distribution.