The above analysis determines whether within cohorts, items from inventories can be combined, that is, whether different inventories can be used to measure the same trait. In addition, it is important to assess whether across cohorts, the same trait is being measured. If Neuroticism and Extraversion were very differently expressed across cohorts, a meta-analysis is rather meaningless. Due to a host of reasons (culture, language, sample selection criteria, etc.), the same test items might have different parameters across cohorts. Ignoring these differences results in systematic bias when comparing individual sum scores from different cohorts. The assumption of equal item parameters across groups is usually termed measurement invariance (Meredith 1993). If one item has different parameter values across groups, this is called differential item functioning (DIF) (Glas 1998, 2001; Speliotes et al. 2010). There are two ways of dealing with DIF, either (1) omitting the item entirely in estimating individual scores, or (2) allowing for different item parameters for that particular DIF item across groups (Weisscher et al. 2010). The first approach leads to loss of information, so that the second is generally more attractive.
