重复自我报告测量带来的数据质量挑战
Intensive, repeated self-report measures are an important tool for behavioral and medical researchers and practitioners who are concerned with the dynamic interplay among variables at a granular level. Many mobile health applications rely on accurate measurement of immediate states and environments for both assessment and intervention delivery. Techniques for capturing repeated momentary assessments yield data with several salutary qualities: recall bias is minimized relative to assessments that rely on much longer recall periods; measurements are taken in individuals’ everyday environments; and dense, repeated measures allow a new window into the processes transpiring between individuals and their environments. In this paper, we highlight several features of repeatedly completing momentary assessments that may change the nature or quality of the data collected over time. Several lines of inquiry are discussed that call into question the presumption that there is invariance in how people complete repeated assessments over time. A result of this possibility could be a reduction in data quality. We present 4 phenomena, with selected results, that may induce noninvariance in repeated measures: the amount of time required to complete assessments, the rate of missing data, the degree of careless responding, and the presence of several components of reactivity. In each of these areas, we found evidence that changes could occur over time, and we consider how data might be affected by such changes. Our conclusion is that researchers should be aware that changes can occur over time and that these changes may affect data quality.密集且重复的自我报告测量方法，是行为学与医学领域研究者和从业者的重要工具，可助力其从精细化层面探究变量间的动态相互作用。诸多移动健康应用在开展评估与实施干预时，均依赖对个体即时状态和所处环境的精准测量。
获取重复性即时评估数据的相关技术，能让所得数据具备多项优良特性：相较于依赖长时回忆的评估方式，该方法可最大限度减少回忆偏倚；测量在个体的日常生活环境中开展，更贴合实际场景；而高密度的重复测量，也为探究个体与环境间的动态作用过程提供了全新视角。
本文重点分析了反复完成即时评估这一操作中，可能导致随时间收集的数据在性质或质量上发生改变的若干特征，同时探讨了多个研究方向 —— 这些研究对 “个体完成重复性评估的方式会随时间保持不变” 这一假设提出了质疑，而该假设若不成立，便可能造成数据质量的下降。
文中梳理了四类可能导致重复测量出现非不变性的现象，并附相关代表性研究结果，具体包括：完成评估所需的时长、数据缺失率、随意作答的程度，以及反应性的若干构成要素。在上述每个研究维度中，均有证据表明相关指标会随时间发生变化，本文也分析了此类变化可能对数据产生的影响。
本研究的结论为：研究者应意识到，相关测量特征会随时间发生改变，且这些变化可能对数据质量产生影响。
https://www.sciencedirect.com/org/science/article/pii/S2291522226000057
