染病同胞對資料中,不同的親子基因型組合依其IBD訊息可分為完整訊息、不確定訊息及完全無訊息三種,當存在不確定訊息與完全無訊息的同胞對資料時,使用傳統均值檢定方法有可能造成統計檢定力的降低。為了提高統計檢定力,本研究利用Franke and Ziegler(2005)所提萃取IBD確定訊息的de Finetti法及張(2006)萃取IBD不確定訊息的熵方法得到染病同胞對的IBD合併訊息量,以有效利用不確定訊息與完全無訊息的樣本,並由此整體訊息量建構出以確定與不確定訊息為基礎的二種加權均值檢定統計量來提升檢定力。經由模擬與過去均值檢定統計量比較,本研究之二種加權檢定統計量在型I錯誤率的表現與過去的檢定統計量相當,而檢定力的表現優於過去研究的方法,確實改善過去研究無法更精確利用不同樣本訊息造成檢定力不足的缺點。
A variety of test statistics can be applied to analysis of affected sib-pairs (ASP) data. However, if the data contain ASP with incomplete information, these statistics might perform in power worse than expected. According to the recent studies, it has been shown that weighting by IBD information could increase the statistical power. Thus, in this study we use the de Finetti method, which extracts the complete IBD information proposed by Franke and Ziegler (2005), and the entropy method, which extracts the uncertain IBD information by Chang (2006), to establish a new weighting scheme to gain extra power. Based on the new weighting scheme, we constructed two information-based weighted mean test statistics. We performed simulation studies for evaluating type I error rates and statistical powers of the two weighted mean test approaches. Simulation results showed that the two weighting approaches do increase statistical powers in various genetic models.