A review of using multilevel modeling in e-learning research. (June 2023)

• Record Type: Journal Article
• Authors: Lin, Hung-Ming; Wu, Jiun-Yu; Liang, Jyh-Chong; Lee, Yuan-Hsuan; Huang, Pin-Chi; Kwok, Oi-Man; Tsai, Chin-Chung
• Is Part Of: Computers & education. Volume 198(2023)
• Publication Date: 2023-06
• Subjects: e-learning -- Multilevel modeling -- Hierarchical linear modeling -- HLM -- Repeated measures; Education -- Data processing -- Periodicals; Education -- Periodicals; Computers -- Periodicals; Computer-Assisted Instruction -- Periodicals; Éducation -- Informatique -- Périodiques; Electronic journals; 370.285
• DOI: 10.1016/j.compedu.2023.104762
• Languages: English
• ISSNs: 0360-1315

Abstract

Abstract: Improving e-learning involves various levels of supports. Accordingly, researchers usually adopt complex research designs with a multilevel structure or repeated measurements to capture a heuristic view of learners' perceptions, comprehension, and behavior in e-learning settings. A total of 76 studies with Hierarchical Linear Modeling (HLM) as a multilevel modeling technique in 13 major e-learning journals from January 2000 to September 2022, published in the Web of Science, were reviewed. We assessed the applications of the following key criteria: reasons for using HLM, data characteristics, sample characteristics, model characteristics, variables used in the research, software use, and main technology used in the research. The results revealed that two-level models and random-intercept models are mostly used in multilevel model building. Moreover, most e-learning studies included two-level random intercept models with "students" as sampling units of analysis in Level 1, and "cognitive learning" (i.e., examination score, learning achievement) as the dependent variable in Level 1. Based on our review results, we provide suggestions and potential applications of using multilevel modeling in e-learning studies. Highlights: Hierarchical Linear Modeling in e-learning has significantly increased recently. Two-level and random-intercept models are the most common models in HLMs. Cognitive learning with student as analysis unit is the most common outcome at L1. ICT experience is the top research topic for HLM studies. Researchers should be sensible and responsive to inherent nested data in e-learning.