Bayesian Hierarchical Duration Models and Meta-Analysis

在政策分析的范围内，这一试点项目将评估对事件发生时间数据进行元分析的统计方法和计算算法的现状，规定对目前做法的改进，并初步制定和实施这些改进。 该研究将在统计科学方面取得进展，对参与教育，法律，卫生，就业，住房，贫困，犯罪和经济等政策领域的统计建模和推断的科学家和研究人员具有广泛的兴趣和实际意义。 评估和开发工作将集中在以下方面：1）时间-事件研究的元分析，包括使用完整数据并将统计输出纳入决策理论框架; 2）时间-事件研究的贝叶斯分层建模，引入新的理论和计算算法用于这些模型的应用;以及3）用于实现基于模拟的贝叶斯事件时间模型的随机模拟和计算机算法。 除了为统计科学广泛感兴趣的元分析开发创新的事件发生时间模型外，该项目还将通过与学科政策分析人员合作，并通过让统计专业的学生进入这些项目领域，促进跨学科应用的进展，从而通过研究和教育促进跨学科的发展。

Within the context of policy analysis, this pilot project will assess the current state of statistical methods and computational algorithms for meta-analysis of time-to-event data, prescribe improvements for current practice, and carry-out preliminary development and implementation of these improvements. The research will make advances in statistical science that are of broad interest and practical relevance to scientists and researchers involved in statistical modeling and inference in the policy areas of education, law, health, employment, housing, poverty, crime, and economics. The assessment and development work will focus on the following: 1) Meta-analysis of time-to-event studies, including use of complete data and incorporating statistical output into a decision-theoretic framework; 2) Bayesian hierarchical modeling of time-to-event studies, introducing new theoretical and computational algorithms for the application of these models; and 3) Stochastic simulation and computer algorithms for implementation of simulation-based Bayesian time-to-event models. In addition to developing innovative time-to-event models for meta-analysis of broad interest to the statistical sciences, the project will contribute to advances in cross-disciplinary applications through collaborations with disciplinary policy analysts and through initiation of students of statistics into such project areas, thus contributing to cross-disciplinary fertilization via research and education.