can partly be explained by differences in age distribution, birth cohort and inclusion criteria. Note that for the NEO, the variances for Neuroticism are larger than for Extraversion, which is explained by the higher reliability of the Neuroticism scale. This is because in the hierarchical modeling, in order to identify scale, the product of the discrimination parameters was fixed at 1, both for Neuroticism and for Extraversion. Larger variance of the latent trait implies that in case the latent variance was fixed to a constant instead of the discrimination parameters, the discrimination parameters would be higher for Neuroticism than for Extraversion. As these discrimination parameters are used for computing test information (Lord 1980), and therefore reliability, we can conclude that Neuroticism is more reliably assessed than Extraversion.Table 3Estimated means and variances of IRT-based Neuroticism and Extraversion latent scores based on NEO-FFI item data, after taking into account measurement non-invariance across cohortsNeuroticismExtraversionCohortMean (SE)VarianceMean (SE)Variance2. BLSA−0.93 (0.04)0.930.50 (0.03)0.563. CILENTO−0.14 (0.03)0.43−0.15 (0.04)0.254. COGEND−0.45 (0.03)0.690.40 (0.03)0.395. ERF−0.28 (0.02)0.380.06 (0.03)0.236. EGCUT−0.16 (0.03)0.370.04 (0.04)0.117. FINNISH TWINS−0.41 (0.04)0.740.34 (0.03)0.418. HBCS−0.59 (0.04)0.650.13 (0.06)0.3711. LBC1936−0.77 (0.04)1.100.25 (0.03)0.5014. NESDA0.05 (0.04)1.120.03 (0.03)0.6215. NTR−0.69 (0.04)0.880.57 (0.03)0.5517. PAGES−1.02 (0.05)0.740.28 (0.07)0.5018. QIMR adolescents−0.11 (0.03)0.600.68 (0.03)0.4919. QIMR adults−0.43 (0.03)0.810.36 (0.03)0.4023. YOUNG FINNS−0.73 (0.04)1.240.50 (0.03)0.61Overall average−0.47
