基于部分核实数据的统计推断及应用

项目名称： 基于部分核实数据的统计推断及应用

项目编号： No.11471060

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 数理科学和化学

项目作者： 邱世芳

作者单位： 重庆理工大学

项目金额： 65万元

中文摘要： 在生物医学研究中，筛检方法价格便宜但其分类误差会导致疾病流行率的有偏估计；金标准方法虽无误差但价格昂贵或有副作用，因此对部分个体同时使用筛检和金标准得到部分核实数据是常用方法。本项目基于部分核实数据对有误判数据探索其统计分析理论和方法，即考虑两种处理方法基于风险差的严格意义上的等价性评价以及在给定容忍限内的等价性评价和非劣性评价，给出大样本渐近的检验过程和小样本精确或近似非条件的检验过程，探索出计算精确或近似非条件置信区间上下限及样本量的有效方法；在分层试验设计下研究基于疾病流行率的等价性评价方法；基于顺序效应度探索在有序误分类数据下两种检验的等价性评价的新方法；在不完全无误判金标准下研究基于部分核实数据的统计分析理论和方法。本项目将在基于部分核实数据的统计推断的研究中取得一系列的理论成果和分析方法，以用于临床医学和流行病学的病理分析与防治中，为医院、药厂和政府获取及时准确的信息提供帮助。

中文关键词： 部分核实数据；缺失数据；贝叶斯方法；顺序效应度；不完全无误判“金标准”

英文摘要： In biometrical studies, screening tests are cheap but the use of data with misclassification could lead to biased estimates. A gold-standard test that is not subject to misclassification, but costly and in some cases intrusive to the test subjects. As a compromise between these two approaches, it is a common strategy to collect the partially validated data that obtained by both using the screening test and the gold-standard test for some of the subjects. Statistical theory and methods will be developed to analyse the categorical and ordinal data with misclassification based on the partially validated series. Fisrt, based on the risk difference for two treatments , we will consider the equivalence test in the strict sense, and then the equivalence test and the non inferiority test with a given tolerance, the asymptotic test procedures under large sample size design and the exact or approximate unconditional test procedures under small sample size scheme will be proposed, some efficient approaches and fast algorithm for calculating the upper and lower confidence limits and accurate sample size will be investigated in this study. Second, the stratified equivalence test based on the disease prevalence will be considered to find the influence factors for a disease. Third, based on the ordinal categorical data with misclassification, we will investigate the equivalence evaluation based on an ordinal effect size measure for two different treatments,and a new method for this equivalence evaluation will be developed. Last, we further consider the statistical theory and methods for the partially validated series with imperfect gold standard. Some important statistical theory and methods for biometrical studies will be obtained from the research of this project, and the statistical methods can be applied to clinical trials or medical research, which will be helpful for hospitals, pharmaceutical factories and governments to obtain the accurate and timely information.

英文关键词： partially validated data;missing data;Bayesian method;ordinal effect size measure;imperfect gold standard