In the 2,613 frequent users, the final model (Table 3) included 12 symptoms. The fit statistics indicated that a two-factor solution had the best model fit (x2=35.12, df=23, p=0.17; RMSEA=0.01). The eigenvalues over 1.0 also indicated two factors; the first factor, termed “Weakness,” and the second termed “Anxiety/Depression.” The item loadings on the factors indicated that the symptoms within each factor comprised a homogeneous set that measured the factor well, with the Weakness factor measured by weakness, hypersomnia, psychomotor retardation, and yawning (factor loadings 0.54–0.99) and the Anxiety/Depression factor measured by anxiety, depression, sweating, nausea, muscle aches, restlessness, tremors and insomnia (factor loadings 0.52–0.88). The two factors were correlated (0.64). In the cannabis-only subsample of 1,119, a similar two-factor model (Table 3) was obtained (x2=15.3, df=16, p=0.50, RMSEA = 0.046). The Weakness factor was measured by the same symptoms found in the larger group (factor loadings 0.65–0.88), and the Anxiety/Depression factor was measured by the same symptoms excepting nausea, which was not retained (factor loadings 0.55–1.00). The two factors were correlated (0.59).
