Differences in the measurements of alcohol use and antisocial behavior items across time may have affected the relationship between the 2 behaviors observed in this study. We aligned the scales of factors measured by different items across time by constraining the measurement parameters to be equal for items that overlapped across time. This method permits the estimation of factors equivalent across time when measurement invariance holds. Measurement invariance across time can be tested by comparing models with or without equality constraints on measurement parameters if there are enough overlapping items to identify the measurement model (Millsap and Yun-Tein, 2004). Unfortunately, our data did not allow the test of measurement invariance across time due to insufficient number of overlapping items, especially at early ages. Thus, we cannot completely rule out the possibility that the relationship between alcohol use and antisocial behavior was affected by noninvariant measurement of the factors across time.
