ABSTRACT
This study examined the measurement invariance of the positive and negative affect scales in the European Social Survey (ESS) in 2006 and 2012. We employed Multi-Group Confirmatory Factor Analysis with an estimator for ordinal data, allowing us to test threshold invariance, which had not been previously investigated for these scales. A 3-item measure of Positive Affect and a 5-item measure of Negative Affect showed that configural, threshold and metric (loading) and partial scalar (intercept) invariance held across almost all countries and between the two ESS Rounds. Our results provide cross-cultural validity to a broader measure of negative affect than past research using the ESS and examine these scales across more countries than any past study. Besides providing valuable insights for researchers interested in well-being and the ESS, our study also contributes to the ongoing discussion about diverging analytical choices in invariance testing.
ABSTRACT
With the increasing availability of cross-national data, more attention has been given to the issue of comparability. But while a lot of emphasis has been directed to the assessment of measurement invariance, there has been substantially less concern on how measurement error can affect the results of measurement invariance testing. In this study, we show how correction for measurement error can be applied to measurement invariance analysis. We illustrate this using the concept of "Perceived ethnic threat" measured in the European Social Survey Round 3 (2006). The measurement invariance results before and after correction for measurement error are compared. We show that correction for measurement error offers a viable way to ensure that non-invariant parameters are actually caused by differences in the data and not caused by the measurement method.
