The choice for the above approach to assessing model fit, which is a bit unconventional, was motivated by the fact that the personality inventories are well-developed and validated instruments. Also, from previous research we know that two-parameter models generally are more appropriate for personality data than one- and three-parameter models (Chernyshenko et al. 2001; Reise and Waller 1990). As one aim is to use as much information as possible from the personality inventories, to establish a linear relationship between personality scales and an external variable, such as a SNP, we chose to retain all items in the analyses.
