The cross-lagged model that is used in our study is illustrated in a path diagram in Fig. 1 for 2 time points. Factors for alcohol use (ALC) and antisocial behavior (ASB) were measured by multiple items at each occasion. The cross-lagged model estimates cross-lagged paths, from 1 factor measured earlier to another factor measured later, as well as autoregressive paths, from a factor measured earlier to the same factor measured later. Cross-lagged paths then represent the effects of a preceding factor to another subsequent factor after accounting for the autoregressive effect on itself. In cross-lagged analyses, it is desirable to have equivalently measured factors across time for more meaningful interpretation of path coefficients. This requirement has been commonly met using the same items over time to measure latent variables and setting their measurement parameters equal across time (White et al., 1993; Young et al., 2008). However, this strategy is not applicable to the ALSPAC data because the measures of alcohol use and antisocial behavior varied over time. Accordingly, we used the item response theory approach to calibrate alcohol use and
