Developing Dynamic Tools for Analyzing Irregularly Spaced Longitudinal Affect Data

情绪的调节是日常生活中不可或缺的一部分。尽管人们对情绪有广泛的兴趣，但很少有统计模型可以正式评估情绪调节的动态。这个项目产生的合作工作将通过引入适合测试现有情绪理论的新方法，促进动态系统建模领域的理论和方法发展。它还将为方法学家创造机会，以改进现有技术并开发新的技术，用于研究动态系统，从而使其他科学学科受益。该项目开发的工具可用于检查在家庭动态、社会网络和疾病传播等其他动态过程研究中经常观察到的不规则间隔的调查数据。医生和其他临床从业者可能受益于有关情绪如何随时间变化的研究结果。为了扩大该项目的教育影响，研究生将参与该项目的所有阶段。这项研究开发的统计工具还将通过在线教程、论坛和会议工作室传播给更广泛的研究受众。对人类动态过程的经验性研究，如对昼夜节律、情绪、疾病传播以及二元和家庭层面互动的研究，经常涉及不规则间隔的纵向调查数据。在对人类情绪的研究中，研究人员经常采用生态瞬时评估(EMA)程序来获得随机或事件应急时间间隔的响应。这样的设计有助于收集反映个人“当下”情绪状态的数据。基于常微分方程和随机微分方程的常用方法可以用来适应在这类数据中观察到的不规则时间间隔，但它们不直接适合于处理具有噪声、高维性质和不同时间尺度特征的EMA数据。不同的时间间隔范围也导致了在确定将微分方程模型与经验数据拟合的适当内插间隔方面的计算挑战。这项研究将产生一套集成的软件工具，用于：(1)拟合和评估连续时间制度转换模型，以提取具有同质动态结构的情绪过程的关键阶段；(2)进行贝叶斯局部影响分析，以评估所提出的建模扩展对假设的先验、样本分布和数据的扰动的敏感性；以及(3)在反映真实EMA研究的条件下，确定用于拟合具有制度转换特征的连续时间模型的最稳健内插间隔。将使用模拟研究以及现有的两个EMA数据集来测试和验证本研究中开发的技术。

The regulation of emotions is an integral part of everyday life. Despite the widespread interest in emotions, very few statistical models exist for formally evaluating the dynamics of emotion regulation. The collaborative work stemming from this project will enhance theoretical and methodological developments in the field of dynamic systems modeling by introducing novel methodologies suited for testing existing theories of emotions. It also will create opportunities for methodologists to refine existing techniques and develop new ones for studying dynamic systems that will benefit other scientific disciplines. The tools developed in this project can be used to examine irregularly spaced survey data frequently observed in the studies of other dynamic processes, such as family dynamics, social networks, and the propagation of diseases. Physicians and other clinical practitioners may benefit from findings concerning how emotions vary over time. To broaden the educational impact of this project, graduate students will be involved in all phases of the project. Statistical tools developed in this study also will be disseminated to a broader research audience through online tutorial, forums, and conference workshops.Empirical studies of human dynamic processes, such as studies of circadian rhythms, emotions, propagation of diseases, and dyadic and family-level interactions, frequently involve irregularly spaced longitudinal survey data. In the study of human emotions, researchers often adopt ecological momentary assessment (EMA) procedures to obtain responses at random or event-contingent time intervals. Such designs facilitate the collection of data that reflect an individual's ongoing emotional states "in the moment." Common approaches based on ordinary and stochastic differential equations can be used to accommodate the irregular time intervals observed in such data, but they are not directly suited for handling the noisy, high-dimensional nature and diverse time scales characterizing EMA data. The diverse range of time intervals also leads to computational challenges in determining the appropriate interpolation intervals in fitting differential equation models to empirical data. The study will yield an integrated set of software tools for: (1) fitting and evaluating continuous-time regime-switching models for extracting key phases of emotion processes with homogeneous dynamical structures; (2) conducting Bayesian local influence analysis to assess the sensitivity of the proposed modeling extensions to perturbations to the hypothesized prior, sampling distribution, and data; and (3) determining the most robust interpolation intervals for fitting continuous-time models with regime-switching features under conditions that mirror real-life EMA studies. Simulation studies as well as two existing EMA data sets will be used to test and validate the techniques developed in this study.