A new alternative Bayesian method for modeling measurement non-invariance (Verhagen and Fox 2013a, b) was applied to assess variance of item parameters across cohorts and that identifies true differences in means and variances of Neuroticism and Extraversion across cohorts, while controlling for any measurement non-invariance. The Bayesian approach allows for estimating complicated models in a straightforward way, and through hierarchical modeling one borrows statistical strength for small cohorts from information in larger cohorts. The Bayesian hierarchical approach assumes there is at least some violation of measurement invariance, and quantifies its extent. Since there are some important differences across cohorts in terms of population and language, we expect there will be at least some difference in item parameters across cohorts.
