Several additional steps were taken to further evaluate the lack of a zero-order correlation between IMPULSE and PEER in Table 2 (where r=.06, p=.08) and the absence of a direct path between these two variables in Figure 3. First, analyses were run separately for adolescents and adults rather than using the combined adolescent and adult sample z-score values in Table 2. This resulted in no significant correlations between impulsivity and PEER. Next, while no Barratt subscale differences regarding relationships with PEER had been predicted, we looked separately at scores for non-planning, motor, and cognitive subscales in adolescent and adult Barratt measures. The results for adults indicated significant correlations with PEER for motor (r=.08, p=.04) and cognitive (r=.08, p=.03) subscales, and in adolescents for the cognitive subscale (r=.13, p=.05). However, when we substituted a latent variable for the Barrett generated from the three Barrett subscales into the SEM in Figure 3, the relationship of IMPULSE and PEER remained nonsignificant within the model (beta weight=.03, p=.50).
