fit for alcohol (Δχ2(1) = 1.39) or cannabis (Δχ2(1) = 2.57), but a deterioration in model fit was observed for cigarettes when the A pathway was dropped (Δχ2(1) = 5.66). The A and C pathways were dropped simultaneously to determine whether an E model best captured the variance in alcohol and cannabis phenotypes, but in both cases, it was a poorer fit for the data (Δχ2(1) = 20.39 and 6.81 for alcohol and cannabis, respectively). Thus, tests of model fit did not support a reduced model for either alcohol or cannabis. The wide range in confidence intervals for A and C values for cigarettes similarly indicated that a reduced model did not provide the best fit for the cigarette use initiation phenotype.
