具有有序依从个体临床试验中因果效应的Bayesian分析  

Bayesian analysis of the causal effect in clinical trials with ordered compliance

作　　者：刘鸿翔[1] 耿直[2] 

机构地区：[1]湖北教育学院数学系,湖北武汉430060 [2]北京大学概率统计系,北京100871

出　　处：《湖北大学学报（自然科学版）》2007年第2期131-137,共7页Journal of Hubei University：Natural Science

摘　　要：在随机处理──对照试验中,经常出现不依从或部分依从的现象,而试验中的受试者所暴露的依从水平往往与自己的身体素质有关.一种情况是身体素质好的受试者可能相信自己对疾病的抵抗能力强而不服药或少服药,也就是说,身体素质好的人,其依从性较差,而身体素质差的人所暴露的依从性却较好;另一种情况则相反,即身体素质好的人其依从性较好,而身体素质差的人所暴露的依从性较差.在单调剂量响应的假设下,通过定义最低有效剂量矩阵,结合上述情况建立因果效应模型,用Bayesian方法对有关的因果效应进行统计推断.In randomized controlled trials, non-compliance or partial compliance of the subjects may often occur, yet their compliance is frequently related to their health conditions. In one case, the subjects who are in better health conditions may probably believe in their own ability to resist any disease without medication or less medication. In other words, this group of people are less compliant, while those in poor health conditions may be more compliant. The other case is on the contrary. Those who are physically strong are more compliant, whereas those in poor health conditions are less compliant. This article, based on the hypothesis of the response of monotone dosage, through defining the lowest possible dosage matrix, has established a mode of causal effect in connection with the above cases and made statistical conclusion on the related causal effect with Bayesian method.

关 键 词：因果效应 依从性 潜在变量 最低有效剂量矩阵 

分 类 号：O211.4[理学—概率论与数理统计]