TPC Journal V8, Issue 2 - FULL ISSUE

182 The Professional Counselor | Volume 8, Issue 2 Table 2 Principal Factor Analysis Results Using Oblique Rotation: Faculty Members (N = 201) Factor 1 (E) Factor 2 (F) Factor 3 (K) Items Loadings h 2 Item #3 0.75 0.58 Item #8 0.68 0.57 Item #10 0.68 0.58 Item #7 0.65 -0.13 0.51 Item #12 0.63 0.38 Item #4 0.86 0.55 Item #9 0.81 0.53 Item #6 0.12 0.67 0.56 Item #2 0.12 0.66 0.57 Item #11 0.65 0.45 Item #1 0.18 0.53 0.33 Item #5 -0.18 0.51 0.38 Eigenvalues 4.74 1.89 1.78 % of variance 32.0 12.5 11.9 Alpha coefficient 0.84 0.83 0.75 Note . Factor loadings over 0.40 appear in bold and mark the particular factor. Blank cells indicate factor loadings ≤ 0.10. E = Engagement; F = Fear; K = Knowledge. Item and internal consistency reliability analyses were computed for the three derived factors to partially answer research question 1. Adequate reliability coefficients were found for the overall measure ( α = .81 ) and for each dimension: engagement (α = .84), fear (α = .83), and knowledge (α = .75). The low correlations between factors (engagement and fear, r = 0.09; engagement and knowledge, r = 0.37; and fear and knowledge, r = 0.11) supported the discriminant validity of the measure. Phase 2: Confirmatory Factor Analysis To cross-validate the CMHPCS with a sample of undergraduate students, a CFA was computed (research question 2). The assumptions necessary for conducting a CFA were met (Byrne, 2016). Multicollinearity was not present, as bivariate correlations did not exceed an absolute value of 0.36. In addition, Mahalanobis d 2 indices revealed no extreme multivariate outliers. The standardized path model is depicted in Figure 1. It was not surprising that the CMIN absolute fit index was statistically significant due to the large sample size: χ 2 (51) = 1.97, p = .007. However, fit indices that are more appropriate for sample sizes larger than 200 revealed an adequate model fit. For example, the CFI = .96, RMSEA = .05, 90% CI [.04, .07], SRMR = .04, and GFI = .95. The path coefficients (see Figure 1) between the engagement and knowledge scales (.48) indicated a stronger relationship than the engagement and fear (.05) or fear and knowledge scales (.07). (These path coefficients are interpreted in the discussion section). Taken together, the CFA results produced a moderate-to-strong fit based on the guidelines from structural equation modeling researchers (Byrne, 2016). Reliability of the dimensions was re-examined with the student sample, yielding similar estimates to those found