- Title
- Multilevel structural equation modelling of complex educational data: investigation of the relationship between teaching quality and student achievement
- Creator
- Tian, Ran
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2024
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- Improving teaching quality has been an ongoing pursuit for policymakers, researchers and broader government agencies and organisations. Professional development (PD) initiatives directed at teachers are a recognised strategy for improving teaching quality, with the underlying intention to support improved student achievement. Quality Teaching Rounds (QTR) is a well-recognised form of collaborative PD in Australia. QTR participation has demonstrated a statistically significant impact on both teaching practices and student achievement in mathematics. The underlying interconnected relationships between QTR, teaching quality, and student achievement, however, remain unexplored due to limited data that often prevents suitable statistical techniques being employed. A large-scale clustered randomised controlled trial (cRCT) investigating the impact of QTR on teaching quality and student achievement was conducted in 2019 in Australia. Longitudinal data were collected from 222 teachers and 5146 students in Stage 2 from 128 primary schools. Teaching practices were rated in the classroom using a classroom observation framework, the Quality Teaching (QT) model which contains 18 elements within three dimensions, while students were assessed using Progressive Achievement Tests (PATs) in mathematics, reading, and science. The data structure therefore comprised a combination of multilevel structure, multiple intervention groups, longitudinal design, and latent constructs of the QT model to measure teaching quality. This PhD study is based on the cRCT data. It examines the underlying interconnected relationships between QTR, teaching quality, and student achievement simultaneously by exploring linear mixed-effects modelling (LMM), structural equation modelling (SEM), and multilevel SEM (MSEM), to incorporate the different aspects of the complex data structure appropriately, providing a valuable framework for future work. Previous studies using the QT model employed the mean of 18 elements for statistical purposes. Converting 18 elements to a single composite measure, however, results in considerable information loss, while measurement error, which commonly exists in classroom observation frameworks, is completely disregarded with LMM. Furthermore, teaching quality, which cannot be directly observed from a classroom observation framework, is typically considered a multidimensional latent construct. This thesis establishes valid and reliable three-factor latent constructs for the QT model via in-depth exploratory and confirmatory factor analysis. This is the first time that valid and reliable latent constructs have been established in the QTR context. Moreover, the comprehensive factor analysis procedures contribute to redefining good practices of measurement modelling in teaching effectiveness studies, which is of significant benefit, given that well-established latent constructs of a classroom observation framework are currently limited. The underlying interconnected relationships are further explored by testing the impacts of the QTR intervention on student achievement in mathematics was mediated by changes in teaching quality via the development, implementation, and comparison of four mediation models that have not been previously examined. Teacher-level student mathematics scores were found to be statistically significantly higher by 0.11SD [95% MCCI = 0.01,0.20] through QTR PD than that through business-as-usual PD in the schools sampled, on average, mediated by the Intellectual Quality dimension of the latent constructs of the QT model. This is the first time that a statistically significant mediation effect has been detected in the QTR context, providing rigorous empirical support for teaching effectiveness studies, where similar studies are limited. Furthermore, a multilevel structure and latent constructs to measure teaching quality were simultaneously incorporated into an integrated model for the first time in the QTR context and in teaching effectiveness studies.
- Subject
- multilevel structural equation modelling; multilevel mediation analysis; educational data with a complex data structure; cluster randomised controlled trial
- Identifier
- http://hdl.handle.net/1959.13/1509982
- Identifier
- uon:56333
- Rights
- Copyright 2024 Ran Tian
- Language
- eng
- Full Text
- Hits: 480
- Visitors: 514
- Downloads: 51
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | ATTACHMENT01 | Thesis | 10 MB | Adobe Acrobat PDF | View Details Download | ||
| View Details Download | ATTACHMENT02 | Abstract | 241 KB | Adobe Acrobat PDF | View Details Download |