In modeling repeated measurements of externalizing problems across nine assessments, an initial comparison was made between traditional linear and quadratic growth curve models to determine whether we should allow for non-linear change in subsequent growth mixture models. Indices of model fit suggested that a model incorporating non-linear growth (χ2, 36 = 229.20; CFI = .90; BIC = 20146.90; AIC = 20073.22) was more representative of the data than a model restricted to linear change (χ2, 40 = 310.49; CFI = .87; BIC = 20203.81; AIC = 20146.51), and a formal likelihood ratio test of the difference in the nested models supported this (χ2Δ, 4 = 81.29, p < .001).
