Figure 1 presents the mean for frequency of HED across the age range by sex, separately for EA and AA groups. Results showed quadratic growth across development, with sex differences. Overall, male participants showed higher levels of HED across time, and more rapid escalation. Both male and female participants showed the normative reduction in HED in the late 20s. Table 2 provides the comparisons of model fit for various multilevel growth models for HED. Results from model fitting supported the observations from visual inspection, with the inclusion of linear and quadratic terms improving overall model fit. In addition, we estimated random effects using restricted maximum likelihood (REML). Each model improved on the fit from the previous. Thus, for all subsequent focal analyses, we fit all models with a random intercept and random slopes for age and age2.Table 2Models for Determining Trajectories Shape and Random Effects StructureFixed effects model fittingAICLLdfConstantAgeAge2Unconditional model46,547.2–23,270.601.408——Linear growth model45,318.1–22,655.110.4860.106—Quadratic growth model44,281.9–22,135.92–0.4310.366–0.013Random effects model fittingAICLLΔdfΔ χ2pRandom intercept44,282–22,136———Random slope for age43,533–21,7592752.97<.001Random slope for age242,863–21,4223675.83<.001Note: Boldface type indicates estimate p < .05. AIC = Akaike information criterion; LL = log likelihood.
