生物医学研究中列联表数据在逆抽样条件下的统计推断

项目名称： 生物医学研究中列联表数据在逆抽样条件下的统计推断

项目编号： No.11461083

项目类型： 地区科学基金项目

立项/批准年度： 2015

项目学科： 数理科学和化学

项目作者： 江绍萍

作者单位： 云南民族大学

项目金额： 36万元

中文摘要： 基于2×2列联表的统计推断研究是目前国内外生物医学和流行病学研究的热门课题。在这些研究中，经常遇到小概率问题。在这种情况下，普通抽样方式得到的观测病例会很少甚至为零，为保证观测到足够病例数可采用逆抽样方法。鉴于此，本项目以逆抽样条件下配对实验设计2×2列联表为研究对象，对优势比的区间假设、齐性检验、等价性评价和置信区间估计等问题进行研究。讨论检验的第一类错误和功效、置信区间的覆盖概率和覆盖宽度及样本量决定问题。研究相应的统计模拟算法和数值计算方法，并尝试将所得理论和分析方法转化到临床医学的实际应用中。相关研究成果将丰富并发展逆抽样条件下列联表数据分析的理论和方法，为复杂分类数据分析提供技术参考，进一步促进生物医学统计应用水平的提升。

中文关键词： 逆抽样；优势比；多重假设检验；配对实验设计；统计推断

英文摘要： The statistical inference of 2×2 contingence table focused on biomedical and epidemiological is a hot research topic in present. The small probability problem is usually existed in these researches. However, a general sampling method can not meet the needs of research because the number of interesting case will tend to zero. The inverse sampling will be used in order to ensure that we can get enough interesting case. For this reason, we plan to perform a systemic research on the matched-pair design under inverse sampling. Firstly, the compound interval hypothesis about risk ratio, homogeneity test of risk ratio, equivalency evaluation and confidence intervals estimation are considered. Then we will calculate the type I error rates and actual power, expected coverage probability (ECP) and expected coverage width (ECW), followed with discussion of the sample size determination problem. Finally，the corresponding statistical simulation algorithm and numerical calculation will be studied. Meanwhile, the theory and analysis methods developed in this study will be proved in practical application. The results will not only enrich and develop the theory and method about the statistical inference of contingence table data under inverse sampling, but also provide technical reference for complex categorical data processing and promote the application level of biomedical statistics.

英文关键词： inverse sampling;odds ratio;multiple hypothesis testing;matched-pair design;statistical inference