28 Apr Multi-Group Structural Equation Models
1:00 pm - 2:30 pm
statistics seminar | level: intermediate |
for questions related to this event, contact firstname.lastname@example.org
affiliation: Ghent University
A fundamental problem in multi-group measurement models is that across groups, the interpretation of questionnaire items can sometimes change. To make comparisons , we want the measurements to be invariant (equivalent) to ensure that the nature of the construct has not changed substantially across groups.
In this seminar, you will be introduced to multi-Group SEM which allows one to test separate structural models in two or more groups or waves. We are going to test for: Configural Invariance (structural equivalence) which assumes that same model holds for all the groups (or waves) i.e. the number of factors and indicator-factor patterns are the same; Metric Invariance (measurement unit equivalence): assumes that factor loadings (slopes) are the same across the groups (or waves); Scalar Invariance (Full score equivalence): the intercepts are the same across all groups or waves being compared; and Strict Factorial Invariance: error variances are the same across all groups Invariance is going to be established using the change is Chi-Square test or change in CFI test. Once Invariance cannot be established, the next step will be to test for partial Invariance.
We are going to use the the classic Holzinger and Swineford dataset (which consists of mental ability test scores of seventh and eighth-grade children) in the lavaan package to test for Measurement Invariance using School (Pasteur and Grant-White) as the grouping variable.
Basic knowledge about SEM
A. Alexander Beaujean. Latent Variable Modeling using R: A Step-By-Step Guide
Kline, R. B. Principles and practice of structural equation modeling, 4th ed. Citation. (2016). (4th ed.). Guilford Press
dr. Emmanuel Abatih