Among those included in the initial GMM, 90 of a possible 512 (i.e., 29) missing data patterns were observed. To summarize, 47.9% had data for all nine externalizing assessments, 44.2% had between four and eight data points (i.e., surpassing the threshold recommended for non-linear analyses), and 7.9% had between one and three assessments. However, the number of available externalizing assessments was not associated with trajectory class membership, χ2(16) = 22.88, p = .12, nor with age 12 reports of peer group antisocial behavior, F(8, 353) = 1.67, p = .10. Likewise, post-hoc comparisons between those who consented to DNA collection and those who refused revealed only negligible differences in trajectory class membership, χ2(2) = 1.22, p = .54, and peer group antisocial behavior, t(383) = 1.24, p = .22.
