Approximate measurement invariance and longitudinal confirmatory factor analysis: concept and application with panel data

Daniel Seddig, Heinz Leitgöb

Abstract


This article addresses the approximate approach to assess measurement invariance with
(longitudinal) confirmatory factor analysis. Approximate measurement invariance uses
zero-mean, small-variance Bayesian priors to allow minor differences in estimated
parameters across time, while still maintaining comparability of the underlying constructs.
The procedure is illustrated for the first time with panel data on young peoples’ preferences
to maximize pleasure and enjoy life. Results indicate whereas the traditional approach of
exact measurement invariance failed to establish scalar invariance across time and precluded
comparisons of latent means, it was possible to establish approximate scalar invariance.
Based on a monitoring procedure for model fit and convergence, a rather small prior variance
was deemed sufficient to account for minor deviations of cross-time intercept differences
from zero.

Keywords


confirmatory factor analysis; Bayesian structural equation modeling; approximate measurement invariance; panel data

Full Text:

PDF


DOI: http://dx.doi.org/10.18148/srm/2018.v12i1.7210

Copyright (c) 2018 Daniel Seddig, Heinz Leitgöb

Hosted by the Library of the University of Konstanz