Statistical methods for administrative data

作为一名生物统计学家，我开发了新的统计方法来改进研究的设计和分析，并在许多项目上合作。虽然我的提案以真实数据和与疾病聚集性检测、相关数据和管理数据源相关的问题为主题，但我的方法可以应用于科学、工程和工业的许多其他领域。疾病聚集性检测方法是一种特殊类型的空间统计方法，可用于识别疾病病例过多的地理区域(称为聚集性或热点)。这些方法对于卫生当局能够监测人口疾病，以便有针对性地进行进一步调查或干预非常重要。我最近在这方面的工作包括创建一种集群检测方法，该方法可以检测与疾病相关的可疑集群，如急诊科(ED)就诊。我在这一领域的拟议研究将把时间纳入集群检测方法，调整方法以包括管理数据中的错误，并确定比率计算与检测相一致的标准。这项工作将包括代数推导和操作，以及计算机模拟研究。这些新进展对于监测疾病或疾病相关事件的研究人员和管理人员将是重要的，这是最近大流行担忧的一个及时话题。其他类型的相关数据也对其他研究目标感兴趣。当可以从个体获得多个响应时，例如多次ED访问，数据是相关的，并且传统的检验两个变量之间的关联性的方法将不再有效。我建议使用重抽样方法为这种情况提供新的统计检验，这样就不需要严格的分布假设。当多个管理数据源组合在一起时，需要将主题分类为组，我将研究主成分分析。这些目标还将包括代数推导和操作以及计算机模拟研究。我的研究将为评估协会和将个人归类提供新的有效方法，这是许多应用研究项目的关键目标。

As a biostatistician, I develop novel statistical methods to improve the design and analysis of studies and work collaboratively on many projects. While my proposal features topics that are motivated by real data and problems related to disease cluster detection, correlated data, and administrative data sources, my methodology could be applied to many other areas in science, engineering, and industry. Disease cluster detection methods are a special type of spatial statistical method that can be used to identify geographic areas with excess cases of disease (called clusters or hot spots). These methods are important for health authorities to be able to monitor populations for diseases in order to target areas for further investigation or intervention. My recent work in this area includes creating a cluster detection method that detects suspected clusters of disease-related events such as emergency department (ED) visits. My proposed research in this area will incorporate time into the cluster detection methods, adapt methods to include errors in administrative data, and determine criteria when rate calculations would agree with detection. The work will include algebraic derivations and manipulations, and computer simulation studies. These new advances will be important for researchers and administrators conducting surveillance of disease or disease-related events, a timely topic with recent pandemic concerns. Other types of correlated data are also of interest for additional research objectives. When multiple responses can be obtained from individuals, like multiple ED visits, the data are correlated and traditional approaches of testing association between two variables will not be valid. I propose to provide new statistical tests for this situation using a resampling method so that strict distributional assumptions are not required. When multiple administrative data sources are combined, and it is desirable to classify subjects into groups, I will investigate principal component analyses. These objectives will also include algebraic derivations and manipulations and computer simulation studies. My research will provide new and valid approaches for assessing associations and classifying individuals into groups, key goals for many applied research projects.