Recursive Estimation of Time-Varying Parameters in Dynamic Factor Models for Nonstationary Psychological TIme Series

许多心理和生物过程具有随时间变化的统计特征。最好的例子是适应、学习和发展过程。最近的数学研究表明，对具有时变统计特征的过程的统计分析必须基于对单个受试者的密集重复测量，才能获得有效的结果。然而，目前缺乏能够对这些过程进行有效分析的统计技术。在这个项目中，将开发创新的统计建模和估计技术，对具有先验未知时变特征的过程进行有效和可靠的分析。新技术可以并将应用于对单个受试者进行密集的实时重复测量，从而能够高保真地跟踪作为环境和受试者特定条件的瞬时变化的函数的心理和生物过程的关键特征随时间变化的波动。在这个项目中开发的新的建模和估计技术将在大规模的计算机模拟研究中得到验证，并在普遍可用的科学软件中实施。它们将应用于具有时变统计特征的一系列心理和生物过程，包括通过脑电和磁共振成像工具测量的大脑对瞬时刺激的反应，以及个人成熟、学习和发展过程。本项目中开发的新的建模和估计技术的一个潜在的特殊应用领域涉及对疾病过程(如1型糖尿病和哮喘)的特定患者的持续评估和最佳治疗。总之，该项目的成果将首次使有效和可靠的统计评估以及对具有时变统计特征的心理和生物过程的最佳指导成为可能。

Many psychological and biological processes have statistical characteristics which vary in time. Prime examples are adaptive, learning and developmental processes. Recently it has been shown mathematically that statistical analysis of processes with time-varying statistical characteristics has to be based on intensive repeated measurements of single subjects in order to obtain valid results. However, at present statistical techniques which would enable valid analyses of such processes are lacking. In this project innovative statistical modeling and estimation techniques will be developed which yield valid and reliable analyses of processes with a priori unknown time-varying characteristics. The new techniques can and will be applied to intensive repeated measurements of single subjects in real time, thus enabling high-fidelity tracking of the time-dependent fluctuations of key characteristics of psychological and biological processes as functions of momentary changes in environmental and subject-specific conditions. The new modeling and estimation techniques developed in this project will be validated in large scale computer simulation studies and implemented in generally accessible scientific software. They will be applied to a range of psychological and biological processes with time-varying statistical characteristics, including brain responses to transient stimuli measured by means of electroencephalographic and magnetic resonance imaging tools, and individual maturational, learning and developmental processes. A potential special field of application of the new modeling and estimation techniques developed in this project involves patient-specific continuous assessment and optimal treatment of disease processes such as diabetes type 1 and asthma. In sum, the outcomes of this project will for the first time enable valid and reliable statistical assessments and optimal guidance of psychological and biological processes with time-varying statistical characteristics.