Based on the Loney et al. (2007) results for the Minnesota Temperament Inventory (MTI), the current study specified the MTI item-to-latent variable relations as follows: Detachment (item #s 8, 13, 14, 15, 16, 17) and Antisocial (6, 7, 9, 10, 11,12, 19). Since the items are ordinal, the robust weighted least squares estimation procedure (WLSMV) provided by the Mplus modeling software (Muthén & Muthén, 2001) was used for all analyses. In addition, the Mplus routine for sample clustering (analysis type = complex) was employed with family as the clustering variable, given the use of a twin sample. (Notably, there was little difference in the results whether the clustering routine was used or not.) For the current study, there were three sets of analyses. First, the Time 1 (age 17) and Time 2 (age 23) MTI items were each analyzed separately to check the adequacy of fit for the two-factor measurement model at each time point. Model fit was determined using a two-index strategy recommended by Hu and Bentler (1999). The Tucker-Lewis index (TLI) was used to assess incremental fit, and
