Reporting Practice in Multilevel Modeling: A Revisit After 10 Years

Wen Luo, Haoran Li, Eunkyeng Baek, Siqi Chen, Kwok Hap Lam, Brandie Semma

Research output: Contribution to journalArticlepeer-review

27 Scopus citations

Abstract

Multilevel modeling (MLM) is a statistical technique for analyzing clustered data. Despite its long history, the technique and accompanying computer programs are rapidly evolving. Given the complexity of multilevel models, it is crucial for researchers to provide complete and transparent descriptions of the data, statistical analyses, and results. Ten years have passed since the guidelines for reporting multilevel studies were initially published. This study reviewed new advancements in MLM and revisited the reporting practice in MLM in the past decade. A total of 301 articles from 19 journals representing different subdisciplines in education and psychology were included in the systematic review. The results showed improvement in some areas of the reporting practices, such as the number of models tested, centering of predictors, missing data treatment, software, and estimates of variance components. However, poor practices persist in terms of model specification, description of a missing mechanism, power analysis, assumption checking, model comparisons, and effect sizes. Updates on the guidelines for reporting multilevel studies and recommendations for future methodological research in MLM are presented.

Original languageEnglish (US)
Pages (from-to)311-355
Number of pages45
JournalReview of Educational Research
Volume91
Issue number3
DOIs
StatePublished - Jun 2021
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2021 AERA.

Keywords

  • methodological development
  • multilevel modeling
  • reporting practices
  • systematic review

Fingerprint

Dive into the research topics of 'Reporting Practice in Multilevel Modeling: A Revisit After 10 Years'. Together they form a unique fingerprint.

Cite this