测量时间与样本量对参数估计的影响：基于密集追踪数据分析的方法比较

The Influence of Measurement Time and Sample Size on Parameter Estimation: Comparison of Methods Based on Intensive Tracking Data Analysis

摘要: 在心理学实验的进行过程中，被试的情绪、想法、行为和生理功能，往往不是保持不变，而是随时间变化一直波动的。传统的追踪或是横断研究能够对群体中存在的现象进行描述解释，但难以解释个体短时间内心理过程的动态变化。密集追踪是在短时间内对个体进行多次测量的方法，测得的数据更利于探究个体在实验过程中心理动态变化的过程作用机制。近年来，随着科学技术的发展，密集追踪测量的难度降低，目前已经成为心理学研究的一大热点。目前针对密集追踪研究的数据分析主要有传统的多层线性模型(Multilevel Modeling, MLM)的方法，以及新兴的动态结构方程模型(Dynamic Structural Equation Modeling, DSEM)的分析方法。二者均可以方便地对密集追踪数据中的自回归和交叉滞后效应进行建模。但目前尚未有研究探讨过时间点和样本量对于这两个模型参数估计的影响，因此本研究以模拟研究的形式，比较在不同时间点和样本量条件下两个模型参数估计的优劣，为研究者在实际研究中选择和使用模型提出建议。

Abstract: In the process of conducting psychological experiments, the emotions, thoughts, behaviors, and physiological functions of the subjects often do not remain unchanged, but fluctuate continuously over time. Traditional tracking or cross-sectional research can describe and explain phenomena that exist within a group, but it is difficult to explain the dynamic changes in individual psychological processes over a short period of time. Intensive tracking is a method of measuring individuals multiple times in a short period of time, and the measured data is more conducive to exploring the process and mechanism of psychological dynamic changes in individuals during the experimental process. In recent years, with the development of science and technology, the difficulty of intensive tracking and measurement has decreased, and it has become a hot topic in psychological research. At present, the data analysis for intensive tracking research mainly includes traditional Multilevel Modeling (MLM) methods and emerging dynamic structural equation modeling (DSEM) analysis methods. Both can conveniently model autoregressive and cross lagged effects in intensive tracking data. However, there is currently no research exploring the impact of time points and sample size on the parameter estimation of these two models. Therefore, this study compares the advantages and disadvantages of parameter estimation of the two models under different time points and sample size conditions in the form of simulation research, and provides suggestions for researchers to choose and use models in practical research.