Statistical Methods for Complex Life History Studies

拟议的研究的广泛目标是开发新的方法来设计和分析复杂的过程和数据。要解决的具体主题是在医学，公共卫生，遗传学和其他领域的应用程序的动机，其中新类型的研究不断被创建。此外，除了与实验或观察性研究有关的较小数据集外，还可获得非常大的数据库，这使研究的某些方面发生了变化，需要采用创新的统计方法来处理与健康、教育、就业和人类生活的其他方面有关的生活史过程。这些过程的复杂性、影响这些过程的因素以及难以收集关于有代表性的个人群体的详细数据，往往使研究工作变得困难，而有偏见或不完整的数据，如果处理不当，可能导致误导性的推论。例如，关于医疗干预或生活方式或环境暴露对健康的影响的断言和结论往往是由于这些因素。使用大型观测数据库（或“大”数据）进行科学发现和推理也需要调查。这样的数据库往往缺少关键变量，并受到其他数据质量问题的影响，这些问题限制了仅基于它们的推断。通过整合来自不同来源的信息，我们可以得出更可靠的结论。研究主要分为三个方面：（1）生活史研究的设计和分析方法;（2）整合来自多个数据源的信息的方法;（3）两阶段或多阶段研究的设计和分析。这项研究将大大增加研究复杂生命史过程的统计工具，并使用多个数据源来了解复杂的现象。与多伦多大学医学、公共卫生和遗传流行病学研究人员合作（例如Laurent Briollais、Shelley Bull、Lei Sun）和其他人（如纽芬兰纪念大学的Yildiz Yilmaz）将推动与疾病和健康有关的研究方面，与Richard Cook和Peisong Han（滑铁卢）的合作将加强生命史分析和大数据利用方面的工作。

The broad objectives of the proposed research are to develop new methodology for the design and analysis of complex processes and data. The specific topics to be addressed are motivated by applications in medicine, public health, genetics and other areas where new types of studies are constantly being created. In addition, some aspects of research are being transformed through the availability of very large data bases, in addition to the smaller data sets associated with experimental or observational studies.Innovative statistical methods are needed to deal with life history processes associated with health, education, employment and other aspects of human lifetimes. The complexity of such processes, the factors affecting them, and the difficulty in collecting detailed data on representative groups of individuals often makes their study difficult, and biased or incomplete data, if not handled appropriately, can result in misleading inferences. Conflicting claims and conclusions concerning medical interventions or the effects of lifestyle or environmental exposures on health are, for example, often due to such factors. The use of large observational data bases (or “big” data) for scientific discovery and inference also require investigation. Such data bases are often missing key variables and suffer from other data quality problems that limit inference based on them alone. By integrating information from different sources, however, we may be able to reach more reliable conclusions.Research is proposed in three main areas: (i) methods for the design and analysis of life history studies, (ii) methods for integrating information from multiple data sources, and (iii) the design and analysis of two- or multi-phase studies. This research will add significantly to the statistical tools for studying complex life history processes, and for using multiple data sources to learn about complex phenomena. Collaborations with researchers in medicine, public health and genetic epidemiology at the University of Toronto (e.g. Laurent Briollais, Shelley Bull, Lei Sun) and elsewhere (e.g. Yildiz Yilmaz, Memorial University of Newfoundland) will motivate aspects of the research related to disease and health, and collaborations with Richard Cook and with Peisong Han (Waterloo) will enhance the work on life history analysis and on the utilization of big data.