Multilevel Structural Equation Models for the Analysis of Comparative Data on Educational Performance

Abstract

The Programme for International Student Assessment comparative study of reading performance among 15-year-olds is reanalyzed using statistical procedures that allow the full complexity of the data structures to be explored. The article extends existing multilevel factor analysis and structural equation models and shows how this can extract richer information from the data and provide better fits to the data. It shows how these models can be used fully to explore the dimensionality of the data and to provide efficient, single-stage models that avoid the need for multiple imputation procedures. Markov Chain Monte Carlo methodology for parameter estimation is described.

Micro–Macro Multilevel Analysis for Discrete Data

“…More information about the estimation procedures can be found in Fox and Glas (2001), Goldstein, Bonnet, and Rocher (2007), Vermunt (2008), Fox (2010), Palardy and Vermunt (2010), Muthén and Asparouhov (2011), and Varriale and Vermunt (2012). In the present study, the software package Latent GOLD (Vermunt & Magidson, 2013) was used to estimate the models.…”

Comment peut-on utiliser les études comparatives internationales pour doter les politiques éducatives d’informations fiables ?

“…H. Goldstein (2004) a montré comment un modèle « bidimensionnel » (c'est-à-dire construit sur deux facteurs) révèle des différences entre la France et l'Angleterre en ce qui concerne l'habitude d'utiliser les questions à choix multiples (voir aussi H. Goldstein et al, 2007). Le premier facteur est un facteur général et le second tend à faire la différence entre les items à choix multiples et les items à réponses ouvertes.…”

“…Ces deux choix comportent des inconvénients et il faut donc que les résultats en tiennent compte. Pour rendre les choses plus claires, voici un exemple de comparaisons entre la France et l'Angleterre (Goldstein et al, 2007).…”

Rasch measurement: a response to Payanides, Robinson and Tymms

“…Payanides et al also seem to be unaware of more recent generalisations of Rasch and other item response models to include multidimensionality, especially within a multilevel structure (see e.g. Goldstein et al, 2007).…”

“…This is, of course, a position that one can take, but is extremely restrictive. It is also one that can be tested on any given assessment, and as Goldstein et al (2007) demonstrate, can be shown not to hold, at least in some cases, where the Rasch model has been used. I also find it difficult to see any theoretical justification for such invariance to be a desirable property of a measuring instrument.…”