Considerations for analyzing EMA data (Oleson et al., 2021)

Purpose: The analysis of Ecological Momentary Assessment (EMA) data can be difficult to conceptualize due to the complexity of how the data are collected. The goal of this tutorial is to provide an overview of statistical considerations for analyzing observational data arising from EMA studies.Method: EMA data are collected in a variety of ways, complicating the statistical analysis. We focus on fundamental statistical characteristics of the data and general purpose statistical approaches to analyzing EMA data. We implement those statistical approaches using a recent study involving EMA.Results: The linear or generalized linear mixed-model statistical approach can adequately capture the challenges resulting from EMA collected data if properly set up. Additionally, while sample size depends on both the number of participants and the number of survey responses per participant, having more participants is more important than the number of responses per participant.Conclusion: Using modern statistical methods when analyzing EMA data and adequately considering all of the statistical assumptions being used can lead to interesting and important findings when using EMA.Supplemental Material S1. Power for given effect sizes, number of participants, and number of surveys per individual for a two independent groups comparison.Supplemental Material S2. Power for given effect sizes, number of participants, and number of surveys per individual for a paired groups comparison.Oleson, J. J., Jones, M. A., Jorgensen, E. J., & Wu, Y.-H. (2021). Statistical considerations for analyzing Ecological Momentary Assessment data. Journal of Speech, Language, and Hearing Research. Advance online publication. https://doi.org/10.1044/2021_JSLHR-21-00081

研究目的：由于数据收集方式的复杂性，生态瞬时评估（Ecological Momentary Assessment，EMA）数据的分析往往难以形成清晰的研究思路。本教程旨在概述针对EMA研究产生的观测数据开展分析时所需关注的统计学考量要点。
研究方法：EMA数据的收集途径多元多样，为统计分析带来了额外的复杂性。本研究聚焦于EMA数据的核心统计学特征，并介绍用于分析EMA数据的通用统计方法。我们依托一项新近开展的EMA相关研究，对上述统计方法进行了实操演示。
研究结果：若设置合理，线性或广义线性混合模型（linear or generalized linear mixed-model）可充分适配EMA收集数据所带来的分析挑战。此外，尽管样本量同时取决于参与者人数与每名参与者的问卷应答次数，但增加参与者数量相较提升单参与者应答次数更为关键。
研究结论：在分析EMA数据时采用现代统计方法，并充分考量所使用的全部统计学假设，可借助EMA研究获得兼具学术价值与实践意义的重要发现。
补充材料S1：针对两组独立样本比较场景，给定效应量、参与者人数及每名个体的问卷调研次数时的统计效力。
补充材料S2：针对配对样本比较场景，给定效应量、参与者人数及每名个体的问卷调研次数时的统计效力。
Oleson, J. J., Jones, M. A., Jorgensen, E. J., & Wu, Y.-H. (2021). 生态瞬时评估数据的统计分析要点. 《语音、语言与听力研究杂志》(Journal of Speech, Language, and Hearing Research). 网络优先出版. https://doi.org/10.1044/2021_JSLHR-21-00081