2008; van Lier et al., 2007). A second issue of measurement is related to the fact that externalizing behavior was measured via 30 items in each of the first six assessment periods, but only 28 items thereafter. Although the variance remained relatively stable across waves, it is possible that the mean decreases starting around 17 years of age are falsely inflated by the truncated range of raw scores at the tail end of the trajectories. However, post-hoc examination of the proportion of items endorsed at each wave revealed a similar downward trend, again beginning in late adolescence. Lastly, while we did make an effort to restrict our analyses to a more genetically homogeneous subsample, it is important to acknowledge the potential for population stratification within self-defined European American populations, where individuals may still derive from diverse ancestries (Heutink & Oostra, 2002).
