Suppressor Variables and Multilevel Mixture Modelling

A major issue in educational research involves taking into consideration the multilevel nature of the data. Since the late 1980s, attempts have been made to model social science data that conform to a nested structure. Among other models, two-level structural equation modelling or two-level path mod...

Full description

Saved in:
Bibliographic Details
Published inInternational Education Journal Vol. 7; no. 2; pp. 160 - 173
Main Authors Darmawan, I Gusti Ngurah, Keeves, John P
Format Journal Article
LanguageEnglish
Published Shannon Research Press 01.06.2006
Subjects
Achievement Tests
Australia
Correlation
Educational Assessment
Educational Research
Evaluation Methods
Foreign Countries
Interaction
Multicollinearity
Path Analysis
Predictor Variables
Science Achievement
Secondary School Students
Australia
Online AccessGet full text

Cover

More Information
Summary:A major issue in educational research involves taking into consideration the multilevel nature of the data. Since the late 1980s, attempts have been made to model social science data that conform to a nested structure. Among other models, two-level structural equation modelling or two-level path modelling and hierarchical linear modelling are two of the techniques that are commonly employed in analysing multilevel data. Despite their advantages, the two-level path models do not include the estimation of cross-level interaction effects and hierarchical linear models are not designed to take into consideration the indirect effects. In addition, hierarchical linear models might also suffer from multicollinearity that exists among the predictor variables. This paper seeks to investigate other possible models, namely the use of latent constructs, indirect paths, random slopes and random intercepts in a hierarchical model. (Contains 4 tables and 11 figures.)
ISSN:1443-1475