Two additional analyses were carried out to further evaluate the SEM results. First, reflecting a concern that the dependence items of tolerance and setting limits on drinking might affect the outcomes and relationships with SRE scores reported here, the model in Figure 2 was repeated after excluding these two items. The problem score based on 10 items correlated with the original 12-item-based score at .96 (p<.0001), and the subsequent model was virtually identical to Figure 2 regarding significant paths, specific path coefficients, and the proportion of the variance explained.
