DHB: Multilevel Autoregressive Moving Average (ARMA) and Dynamic Models for Longitudinal Data and the Study of Human Interactions

DHB：纵向数据的多级自回归移动平均 (ARMA) 和动态模型以及人类交互的研究

• 批准号: 0527449
• 负责人: Michael Rovine
• 金额: $ 74.94万
• 依托单位: Pennsylvania State Univ University Park
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-02-01 至 2010-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0527449&HistoricalAwards=false
• 关键词: DHB; Multilevel; Autoregressive; Moving; Average

Scientists interested in describing the complexity of human interaction have begun to realize the necessity of collecting intensive longitudinal data to enable them to describe the dynamics of the interaction process. These data can take the form of a daily diary of marital interactions, an hourly account of the behavior of an unruly student, a second to second account of a parent-infant interaction, a continuous time measure of physiological characteristics, or a dose-to-dose account of the effect of a drug or treatment regimen. Unlike the more common approaches to the study of development that may use a single, simple model to describe average group trends, dynamic models can accommodate differences in how individuals change. They can be used to describe how a process unfolds for each individual, to show how individuals differ in that process, and to predict how that process will evolve. Once these dynamics are understood, the opportunity exists to develop a control model with the intent of moving the process toward a desired outcome. As part of this project, the investigators plan to adapt, modify, and when necessary develop new dynamic and control models appropriate for the needs of developmental researchers. To accomplish these goals, this project brings together members of the engineering, applied developmental statistics, and developmental science research communities to develop models that will specifically address questions related to differences in the way individuals develop and interact.To demonstrate this model data from the Infant and Child Temperament Study will be assessed to determine the individual differences in how an infant develops the ability to self-regulate emotion. The state-space implementation of this model will allow more flexibility and a richer description of the dynamics of the process by producing a model that can change and develop as the process changes. This approach leads directly to the optimal control/process adjustment model in which individuals can be assessed to determine whether they moving in the direction of a desired outcome. If they are off target, critical control variable values can be assessed and changed. The results of these changes can then be assessed to determine if the individual is back on target. To demonstrate this model, the investigators will assess parent-infant interaction data related to the parents' ability to soothe a distressed child. A model will be developed that will allow investigators to assess which combinations of parent behavior and infant response result in decreasing the child's distress. This model can then be used to provide online assessment to parents that will give them useful feedback during the course of the subsequent interactions suggesting whether their behavior should lead to success and whether and how they should modify that behavior. This general analytic approach will allow researchers to model many different situations in which such feedback can help optimize the quality of an interaction. These can include marital interactions, parent/child relationships, interactions between teachers and students, and encounters between therapists and clients. These methods will also be appropriate for optimizing individual outcomes based on drug treatments, physical therapies, medical treatments, and combinations of different treatment regimens.

对描述人类互动的复杂性感兴趣的科学家已经开始意识到收集大量纵向数据的必要性，以使他们能够描述互动过程的动态。这些数据可以采取婚姻互动的每日日记、对不守规矩的学生行为的每小时记录、父母与婴儿互动的每秒记录、生理特征的连续时间测量或剂量的形式。 -药物或治疗方案效果的剂量说明。与可能使用单一、简单的模型来描述平均群体趋势的更常见的发展研究方法不同，动态模型可以适应个体变化的差异。它们可用于描述每个个体的过程如何展开，显示个体在该过程中的差异，并预测该过程将如何演变。 一旦理解了这些动态，就有机会开发一个控制模型，旨在使过程朝着期望的结果发展。作为该项目的一部分，研究人员计划适应、修改并在必要时开发适合发展研究人员需求的新动态和控制模型。 为了实现这些目标，该项目汇集了工程、应用发展统计学和发展科学研究界的成员来开发模型，专门解决与个体发展和互动方式差异相关的问题。为了证明该模型的数据来自婴儿将评估儿童气质研究，以确定婴儿如何发展自我调节情绪的能力的个体差异。 该模型的状态空间实现将通过生成一个可以随着过程变化而变化和发展的模型，从而提供更大的灵活性和更丰富的过程动态描述。 这种方法直接导致最佳控制/过程调整模型，在该模型中可以评估个体以确定他们是否朝着期望结果的方向前进。如果它们偏离目标，则可以评估和更改关键控制变量值。 然后可以评估这些变化的结果，以确定个人是否回到目标。 为了证明这个模型，研究人员将评估与父母抚慰痛苦儿童的能力相关的亲子互动数据。将开发一个模型，使研究人员能够评估父母行为和婴儿反应的哪些组合可以减少孩子的痛苦。 然后，该模型可用于向父母提供在线评估，在随后的互动过程中为他们提供有用的反馈，建议他们的行为是否应该导致成功，以及他们是否应该以及如何改变这种行为。这种通用的分析方法将允许研究人员对许多不同的情况进行建模，在这些情况下，此类反馈可以帮助优化交互的质量。 这些可以包括婚姻互动、父母/孩子关系、教师和学生之间的互动以及治疗师和客户之间的接触。这些方法也适用于基于药物治疗、物理治疗、药物治疗以及不同治疗方案的组合来优化个体结果。