Time series models of intraindividual variability

个体内变异性的时间序列模型

• 批准号: 7093739
• 负责人: DANIEL H SPIELER
• 金额: $ 6.54万
• 依托单位: GEORGIA INSTITUTE OF TECHNOLOGY
• 依托单位国家: 美国
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-07-15 至 2008-06-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7093739
• 关键词: age difference; aging; attention; behavioral /social science research tag; clinical research; cognition; data collection methodology /evaluation; human old age (65+); human subject; longitudinal human study; mathematical model; memory; model design /development; psychoanalysis; psychological models; statistics /biometry; time resolved data; young adult human (21-34)

DESCRIPTION (provided by applicant): Measures of cognitive performance are subject to age-related changes. In addition to changes in the level of these measures, there are also consistent increases in intraindividual variability. Understanding the meaning of these age-related changes in intraindividual variability is important both for a basic understanding of cognitive aging and for understanding of the diagnostic utility of variability measures. Theoretical accounts for age-related increases in intraindividual variability often refer to time dependent processes such as fluctuations or lapses of attention. Data in this context are rarely analyzed as time series despite this being the most appropriate form to reveal such processes. We propose the application of time series analyses to explore the sequential relations in performance for young and older adults. We show preliminary evidence that time series data from both young and older adults are long range dependent (LRD). LRD refers to the fact that relations can be observed in time series data spanning many trials. Temporal aggregation of component cognitive processes can produce LRD and the strength of LRD is influenced by the operation of longer time scale processes that include executive control processes that act to maintain performance over periods of time. Thus, LRD may be useful for understanding how these processes are influenced by variations in processing demands and the aging process. However, tools that have been used to model these dependencies called ARFIMA models make assumptions of stationarity that do not hold for behavioral data. We propose SEMIFAR models as better suited to measuring LRD and are robust in the face of nonstationarity of behavioral data. We also propose the application of GARCH/ARCH models that capture sequential relations in variability. Preliminary data suggests the time series for young and older adults show volatility clustering. In other words, high and low variability periods cluster together within an individual's time series. Together, these models provide robust methods for exploring the contribution of time dependent processes to intraindividual variation. Four experiments will be analyzed using these models to account for age differences in these components of intraindividual variabilty. These experiments explore manipulations designed to influence sequential dependencies within young and older adults and extend our understanding of sources of intraindividual variability.

描述（由申请人提供）：认知表现的测量会受到与年龄相关的变化的影响。除了这些措施水平的变化之外，个体内变异性也持续增加。了解这些与年龄相关的个体内变异性变化的含义对于基本了解认知衰老和理解变异性测量的诊断效用都很重要。与年龄相关的个体内变异性增加的理论解释通常涉及时间依赖性过程，例如波动或注意力缺失。尽管这是揭示此类过程的最合适形式，但很少将这种情况下的数据作为时间序列进行分析。我们建议应用时间序列分析来探索年轻人和老年人表现的顺序关系。我们的初步证据表明，年轻人和老年人的时间序列数据都具有长程依赖性（LRD）。 LRD 指的是可以在跨越许多试验的时间序列数据中观察到关系的事实。组件认知过程的时间聚合可以产生 LRD，并且 LRD 的强度受到较长时间尺度过程的运行的影响，这些过程包括在一段时间内维持性能的执行控制过程。因此，LRD 可能有助于理解这些过程如何受到加工需求和老化过程变化的影响。然而，用于对这些依赖性进行建模的工具（称为 ARFIMA 模型）做出了不适用于行为数据的平稳性假设。我们提出 SEMIFAR 模型更适合测量 LRD，并且在面对行为数据的非平稳性时具有鲁棒性。我们还建议应用 GARCH/ARCH 模型来捕获变异性的序列关系。初步数据表明，年轻人和老年人的时间序列显示出波动性聚集。换句话说，高变异性时期和低变异性时期在个体的时间序列中聚集在一起。这些模型共同提供了稳健的方法来探索时间相关过程对个体内变异的贡献。将使用这些模型分析四个实验，以解释个体变异性的这些组成部分的年龄差异。这些实验探索了旨在影响年轻人和老年人的顺序依赖性的操作，并扩展了我们对个体内变异来源的理解。