School of Psychology, Zhejiang Normal University, Jinhua, China.
Intelligent Laboratory of Child and Adolescent Mental Health and Crisis Intervention of Zhejiang Province, Zhejiang Normal University, Jinhua, 321004, China.
Behav Res Methods. 2024 Mar;56(3):1656-1677. doi: 10.3758/s13428-023-02113-5. Epub 2023 Apr 14.
To measure the parallel interactive development of latent ability and processing speed using longitudinal item response accuracy (RA) and longitudinal response time (RT) data, we proposed three longitudinal joint modeling approaches from the structural equation modeling perspective, namely unstructured-covariance-matrix-based longitudinal joint modeling, latent growth curve-based longitudinal joint modeling, and autoregressive cross-lagged longitudinal joint modeling. The proposed modeling approaches can not only provide the developmental trajectories of latent ability and processing speed individually, but also exploit the relationship between the change in latent ability and processing speed through the across-time relationships of these two constructs. The results of two empirical studies indicate that (1) all three models are practically applicable and have highly consistent conclusions in terms of the changes in ability and speed in the analysis of the same data set, and (2) additional analysis of the RT data and acquisition of individual processing speed measurements can reveal the parallel interactive development phenomena that are difficult to detect using RA data alone. Furthermore, the results of our simulation study demonstrate that the proposed Bayesian Markov chain Monte Carlo estimation algorithm can ensure accurate model parameter recovery for all three proposed longitudinal joint models. Finally, the implications of our findings are discussed from the research and practice perspectives.
为了使用纵向项目反应准确性（RA）和纵向反应时间（RT）数据来衡量潜在能力和加工速度的平行交互发展，我们从结构方程模型的角度提出了三种纵向联合建模方法，即基于非结构化协方差矩阵的纵向联合建模、基于潜在增长曲线的纵向联合建模和自回归交叉滞后纵向联合建模。所提出的建模方法不仅可以单独提供潜在能力和加工速度的发展轨迹，还可以通过这两个结构的跨时间关系利用潜在能力变化与加工速度之间的关系。两项实证研究的结果表明：（1）所有三种模型在分析同一数据集的能力和速度变化时都是实用的，并且具有高度一致的结论；（2）对 RT 数据的额外分析和个体加工速度测量的获取可以揭示仅使用 RA 数据难以检测到的平行交互发展现象。此外，我们的模拟研究结果表明，所提出的贝叶斯马尔可夫链蒙特卡罗估计算法可以确保所有三种提出的纵向联合模型的准确模型参数恢复。最后，从研究和实践的角度讨论了我们发现的意义。
