New Statistical Models for Intensive Longitudinal Data

描述（由申请人提供）：技术创新彻底改变了健康研究中的科学研究和知识发现过程，使研究人员能够轻松收集具有许多紧密间隔测量场合的密集纵向数据。密集的纵向数据的收集为更好地理解广泛的身体健康和精神健康状况的出现和临床过程提供了很大的希望。理论上，密集的纵向数据可以为健康研究中的重要问题提供答案。然而，旨在利用丰富的密集纵向数据的统计方法目前落后于这些数据收集能力。例如，目前还不清楚什么样的统计程序可以应用于密集的纵向数据，以解决以下问题：在一天或一周内，戒断的主观感觉是如何变化的？情绪和药物使用之间的关系是什么？这种关系在不同的研究对象之间会发生变化吗？如果人口是几个子群体的组合，那么子群体之间的关系是否不同？在本项目中，我们分别提出了具有连续响应、二元响应和计数响应的密集纵向数据的三种新的统计模型。这些新模型具有许多有价值的特点，使它们最适合用于解决健康研究和药物滥用研究中的关键问题。提出的新模型允许总体由几个子组组成，并且影响随时间和个体主体的变化而变化，并且它们使错误过程的结构非常灵活。我们将为新模型提出评估程序，并开发软件以实施所建议的程序。我们计划应用提出的程序，使用烟草、酒精和大麻的经验数据来检验关于药物使用的重要假设，并使用哮喘的经验数据来解决健康研究中的重要问题。我们还计划在统计和行为/社会科学文献中发表拟议的研究，并通过免费软件向健康研究领域的科学家广泛提供新程序。因此，拟议的研究将为各种健康相关研究领域的科学家提供他们需要的工具，利用密集的纵向数据来解决核心科学问题。拟议的程序将用于利用烟草、酒精和大麻的经验数据来解决核心药物使用问题，这些物质是美国使用最广泛的物质，与无数短期和长期后果有关。拟议的程序还将用于使用哮喘的经验数据来测试健康研究中的重要假设。哮喘是一个主要的健康问题，因为据认为美国有1000万人患有哮喘。

DESCRIPTION (provided by applicant): Technological innovations have revolutionized the process of scientific research and knowledge discovery in health studies and allow researchers to easily collect intensive longitudinal data that have many closely spaced measurement occasions. The collection of intensive longitudinal data holds much promise for better understanding the emergence and clinical course of a wide range of both physical health and mental health conditions. In theory, intensive longitudinal data can provide answers to important questions in health studies. However, statistical methodology designed to capitalize on the richness of intensive longitudinal data currently lags behind these data collection abilities. For instance, it is not immediately clear that what statistical procedures can be applied to intensive longitudinal data to address questions such as: How does the subjective sensation of withdrawal vary over a day or a week? What is the relationship between mood and drug use? Does the relationship change across individual subjects? Does the relationship vary across subgroups if the population is a combination of several subgroups? In this project, we propose three new classes of statistical models for intensive longitudinal data with a continuous response, binary response and count response, respectively. These new models possess many valuable features which make them the most appropriate to use for addressing critical questions in health studies and drug abuse researches. The proposed new models allow populations to be composed of several subgroups, and effects to vary over time and change across individual subjects, and they keep the structure of the error process very flexible. We will propose estimation procedures for the new models, and develop software to implement the proposed procedures. We plan to apply the proposed procedures to test important hypotheses about drug use using empirical data on tobacco, alcohol and marijuana, and address important questions in health studies using empirical data on asthma. We also plan to publish the proposed research in both the statistical and behavioral/social science literature, and make the new procedures widely available to scientists in health studies, by means of software free of charge. Thus, the proposed research will provide scientists in various health-related research areas with tools they need to address central scientific questions using intensive longitudinal data. The proposed procedures will be employed to address central drug use questions using empirical data on tobacco, alcohol and marijuana, which are the most widely used substances within the US and have been linked to a myriad of both short and long-term consequences. The proposed procedures will also be used to test important hypotheses in health studies using empirical data on asthma, which is a major health problem as there are thought to be 10 million people with asthma within the United States.