项目名称: 复杂数据下带有形状约束的半参数模型统计推断
项目编号: No.11471065
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: 王晓光
作者单位: 大连理工大学
项目金额: 60万元
中文摘要: 本项目致力于研究半参数模型统计推断,结合生物医学等领域中对于非参数函数形状的特殊约束条件,解决日益迫切的复杂数据分析问题。缺失数据、删失数据、纵向数据、零膨胀数据等越来越多的出现在生物医学数据分析之中,经典的统计模型和方法无法直接处理这些复杂数据。为了获得准确而有效的数据分析结果,通常还要考虑协变量的参与,此时半参数模型尤为重要。为强调应用,拟研究一类特殊的半参数模型,其中非参数函数呈现出某种特殊的形状约束,如单调递增、凸或凹,这是在研究剂量-反应曲线、艾滋病数据、受试者工作特征曲线(ROC)等实际问题中不可或缺的约束条件。这种条件如果被忽视,所得估计常常是有偏、低效的。这些问题的研究一方面对半参数模型估计方法提出新的挑战,另一方面要灵活的处理各种复杂数据,同时也对算法要求更高。本项目以应用为出发点,结合灵活的半参数模型,希望为在生物医学统计领域中的实际问题提供新方法、新算法及其理论基础。
中文关键词: 半参数模型;多元统计分析;参数估计;非参数估计;统计推断
英文摘要: The investigator studies the statistical inference for the semiparametric model with the shape restrictions for the nonparametric functions to account for the urgent complex data analysis in medicine and biology sciences. The traditional statistical models and methods can't fit the complex data directly, which arise at the biomedical research including missing data, censoring data, longitudinal data, zero-inflated data and etc. The semiparametric model is of great importance when the covariates are involved to obtain more correct and efficient results. A class of special semiparametric models will be studied in response to nonignorable shape restrictions for applications such as dose-response curve, HIV data and ROC curve. Without these shape restriction conditions, the estimation will be biased and inefficient. The proposed models and methods are new challenge for the estimation of semeparametric models and help dealing with all kinds of complex data flexibly with the high need of computational algorithms. This proposal is motivated by biomedical practice and based on semiparametric models. The results contribute to the advancement of the statistical theory on new estimation methods and algorithms.
英文关键词: Semiparametric Model;Multivariate Statistical Analysis;Parametric Estimation;Nonparametric Estimation;Statistical Inference