Optimal parameters setting of flexible strategy testing for subgroup analysis

• Main Authors: Yan-LongPan, 潘彥龍
• Other Authors: Yi-Hsuan Tu
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/12220090372719435388

碩士 === 國立成功大學 === 統計學系 === 102 === Targeted therapy can more precisely identify and attack cancer cells or inhibit important molecular signaling pathway, usually while doing little damage to normal cells and achieving treatment goals. This new treatment regimen has been successfully applied to a number of cancers. However, some studies indicate that targeted therapy has a significant effect only for special groups of patients. It carries out that not only the efficacy of overall population but also that of subgroup analysis is an important issue in targeted therapy clinical trials. In order to reduce the threshold to enter the subgroup analysis, Alosh and Huque (2009) proposed a flexible strategy for testing subgroups and overall population. This strategy introduces new parameters to improve the power of subgroup test effectively while the familywise Type I error rate is controlled. It brings out an open question that whether there exist any optimal parameter settings for a given alternative hypothesis. In this thesis, first, we will discuss the influence of power of overall population test and subgroup test in flexible strategy testing. Second, we develop two criteria in the concept of loss less than obtainment (means the increases of power of subgroup test is less than the decreases of power of overall population test) as a starting point: (1) the maximization of power for subgroup test, and (2) the maximization of power for overall population test and subgroup test. Finally, we provide some results for two criteria.