Sensitivity Analysis on Crossover Design Model in Bioequivalence

• Main Authors: Ko, Cheng-Yen, 柯政言
• Other Authors: Huang, Yu-Fen
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/23948618389668396261

碩士 === 國立中正大學 === 數學系統計科學研究所 === 102 === A crossover design model is a commonly used statistical model for the bioequivalence study. In practice, some outliers may affect the results of the analysis and cause wrong decision in bioequivalence. Therefore, it is important tofind out the influential observations. Chow and Tse (1990) discussed this issue by using the likelihood distance and estimates distance. Huang and Ke (2013) proposed the influence functions to detect the influential subjects for the bioequivalence study. However, the aforementioned methods just consider the effect of subjects in the model. In this thesis, we focus on developing influence functions for the full model of a crossover design experiment. Moreover, the comparison between the proposed approach and the methods proposed by Chow and Tse are investigated. Two simulated data and two real data are provided to illustrate the results of these approaches.