in a cross-lagged framework using the program AMOS 18. A non-significant chi square, Comparative Fit Index (CFI) >.95, and Root Mean Square Error of Approximation (RMSEA) <.08 were considered indices of good fit (Hu & Bentler, 1999; Kline, 2005). As the chi square statistic is known to be sensitive to minor deviations from a perfectly causal model (Byrne, 2001; Schumacker & Lomax, 2004) a χ2/df ratio below 3 was considered acceptable (Carmines & McIver, 1981). Measurement error of each variable at a given time-point was allowed to co-vary with the measurement error of that variable at every other time-point to account for shared variance associated with each indicator (Pitts, West, & Tein, 1996). After deriving an acceptable SEM model with the full sample, we ran a multi-group model to examine differences in longitudinal associations between H/I and TD groups. A fully constrained model with every path held constant across groups was used to formally test significance of difference in each path.
