Collaborative Research: Temporal Configuration Analysis for Extracting Qualitative Information from Multi-Wave, Multi-Dimensional Data

合作研究：从多波、多维数据中提取定性信息的时间配置分析

• 批准号: 0532296
• 负责人: Edward Ip
• 金额: --
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-10-01 至 2008-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0532296&HistoricalAwards=false
• 关键词: Collaborative; Research; Temporal; Configuration; Analysis

The project will develop an analytic methodology that will provide researchers with a means for extracting qualitative information about the dynamics of individual behavior from multi-wave, multi-variable data sets. This innovation will be timely and useful in that it coincides with recent improvements in study design and data collection methodologies that have led to increased availability of and interest in longitudinal data among social, behavioral, and economic researchers. The methodology will be motivated by a diverse range of applications that include studies of crop diversification of tobacco farmers, suicidal behavior of adolescents, and emotional changes monitored through "Experience Sampling Methods," among others. The methodology will be called Temporal Configuration Analysis (TCA) so as to emphasize the focus on detection and identification of meaningful patterns in individual behavior trajectories over time. The key element in TCA will be a dynamic model that tracks the transition of "latent states" from one time point to another for each individual. Latent states (e.g., health state, psychological state, political inclination) can be viewed as summary measures of observed variables that translate to or condition manifested behavior. TCA will allow a researcher to identify homogeneous subgroups of individuals in terms of their temporal trajectories of latent states and then to examine the profile of each subgroup. Accordingly, a researcher will be able to interpret the results from analysis of multi-wave data in terms of both observed and latent measures and their trajectories. This will provide additional behavioral meaning and insight. TCA will also feature a broad range of component statistical models to provide users with flexibility in handling commonly encountered data characteristics, such as cross-sectional dependence, state-dependence, and serial correlation. TCA will create a statistical framework for analyzing multi-wave data in which inter-temporal qualitative information is of interest.The project will have immediate, broad, and significant impact on the social, behavioral, and economic sciences and statistical research. First, researchers will be able to analyze multi-wave data in a new way that is different from and more informative than traditional methods such as latent growth curve analysis. For example, it will be possible to describe and visualize behavioral responses by projecting these configurations onto the space of actual behavior. Second, the project will contribute to the science of statistics, specifically in the area of hidden Markov models (HMM). The investigators expect that once this method is developed, explicated, and applied, statisticians and mathematical scientists will expand the TCA framework to include additional innovative research applications. The project's impact will be broadened and maximized in several ways: (1) dissemination of accessible, user-friendly, high-quality software that will document the TCA methodology and allow researchers to use it easily; (2) integration of research and education at several levels-grades 10-12, undergraduate, and graduate; and (3) interaction between the team of applied and theoretical investigators and various research groups, both nationally and internationally, with emphasis on publishing applied research findings. The project will deliver Web-based end-products-high-quality programs in packages familiar to social, behavioral, and economic researchers. The investigators will collaborate with the Center of Excellence in Research, Teaching, and Learning (CERTL) at Wake Forest University, using its infrastructure and experience in training American students in science and engineering, to integrate research into education. The project team, which includes national and international researchers from several disciplines, will leverage this collaboration to increase both the impact and visibility of the project through extensive coordination with various national and international research groups. This award was supported as part of the fiscal year 2005 Mathematical Sciences priority area special competition on Mathematical Social and Behavioral Sciences (MSBS).

该项目将开发一种分析方法，为研究人员提供一种从多波、多变量数据集中提取关于个体行为动态的定性信息的方法。这一创新将是及时和有用的，因为它与最近研究设计和数据收集方法的改进相吻合，这些改进导致社会、行为和经济研究人员对纵向数据的可用性和兴趣增加。该方法将受到多种应用的推动，包括对烟草种植者作物多样化的研究，青少年的自杀行为，以及通过“体验抽样方法”监测的情绪变化等。该方法将被称为时间配置分析（TCA），以强调随着时间的推移，个人行为轨迹中有意义的模式的检测和识别。TCA的关键要素将是一个动态模型，它可以跟踪每个人从一个时间点到另一个时间点的“潜在状态”的转变。潜在状态（例如，健康状态、心理状态、政治倾向）可被视为对观察到的变量的总结性度量，这些变量转化为或制约了表现出来的行为。TCA将允许研究人员根据潜在状态的时间轨迹识别同质的个体亚群，然后检查每个亚群的概况。因此，研究人员将能够根据观察到的和潜在的测量及其轨迹来解释多波数据分析的结果。这将提供额外的行为意义和洞察力。TCA还将以广泛的组件统计模型为特征，为用户提供处理常见数据特征的灵活性，例如横断面依赖性、状态依赖性和串行相关性。TCA将创建一个统计框架，用于分析跨时间定性信息感兴趣的多波数据。该项目将对社会、行为、经济科学和统计研究产生直接、广泛和重大的影响。首先，研究人员将能够以一种新的方式分析多波数据，这种方法与潜在生长曲线分析等传统方法不同，而且信息量更大。例如，通过将这些配置投射到实际行为的空间中，将有可能描述和可视化行为反应。其次，该项目将有助于统计科学，特别是在隐马尔可夫模型（HMM）领域。研究人员期望，一旦这种方法被开发、阐明和应用，统计学家和数学科学家将扩展TCA框架，以包括更多的创新研究应用。该项目的影响将在以下几个方面得到扩大和最大化：(1)传播可访问的、用户友好的、高质量的软件，这些软件将记录TCA方法，并使研究人员能够轻松使用；(2)在10-12年级、本科生和研究生等多个层次进行研究和教育的整合；(3)应用和理论研究团队与国内外各种研究小组的互动，重点是发表应用研究成果。该项目将以社会、行为和经济研究人员熟悉的方式提供基于网络的最终产品——高质量的程序。研究人员将与维克森林大学（Wake Forest University）的卓越研究、教学和学习中心（CERTL）合作，利用其在培训美国科学和工程学生方面的基础设施和经验，将研究与教育结合起来。项目团队包括来自多个学科的国内和国际研究人员，他们将通过与各个国家和国际研究小组的广泛协调，利用这一合作来增加项目的影响和知名度。该奖项是作为2005财政年度数学科学优先领域数学社会和行为科学特别竞赛的一部分得到支持的。