TPC Journal V7, Issue 3 - FULL ISSUE

The Professional Counselor | Volume 7, Issue 3 229 next step was to examine the normality of the data and determine the most appropriate method of extraction. To assess for normality of our data, we checked the univariate normality of each SOCCS item, and if item univariate normality was satisfied, we checked multivariate normality using the Mardia test (Mvududu & Sink, 2013). We identified several SOCCS items that were not normally distributed; therefore, multivariate normality was not examined because univariate normality is a necessary condition of multivariate normality (Mvududu & Sink, 2013). In addition, our histograms, boxplots, and Q-Q Plots results identified that multiple SOCCS items were non-normally distributed; hence, we assumed the data was non-normally distributed, which can occur in social science research (Mvududu & Sink, 2013). Data Analysis After screening the dataset for missing data and assessing for normality, we conducted an EFA to examine the factor structure of the SOCCS with our sample of counseling practitioners and counselors- in-training. Because of the non-normality of the data (Costello & Osborne, 2005), PAF was used for extraction with an oblimin rotation with Kaiser Normalization. A significant value ( p < .001) was identified for Bartlett’s test of sphericity (Bartlett, 1954), and a value of .83 was obtained for Kaiser- Meyer-Olkin sampling adequacy for the SOCCS. Next, we examined internal consistency reliability of SOCCS using Cronbach’s α, thus assessing the degree of correlation between SOCCS items. Results To examine the factor structure of SOCCS, we used EFA, employing PAF analysis. All SOCCS items displayed a factor loading of at least .3 and were initially retained (Floyd & Widaman, 1995; Hair et al., 2010). However, SOCCS items were reduced following classical test theory in order to reduce items with poor measurement properties and to increase internal consistency reliability (Crocker & Algina, 2006; DeVellis, 2003). As noted in Table 2, The PAF results identified the presence of six SOCCS factors with eigenvalues exceeding one, explaining 62% of the variance. However, the first three factors produced eigenvalues of greater than 2.8, whereas the remaining three were all less than 1.5. The three factors accounted for 49% of the variance. As noted in Figure 1, the scree plot, a preferred method for identifying factor solutions in EFA (Hair et al., 2010), identified a steep decline including three factors, a break near the fourth factor, and a significant plateau at the fifth factor, supporting a 3- or 4-factor model solution for the SOCCS with these data. The factor matrix showed loadings of more than .4 for the first three factors, and less than .4 for the fourth through sixth factors. The first three SOCCS factors paralleled Bidell’s conceptually based factors of Skills, Awareness, and Knowledge. In the essence of EFA, we examined the potential construct being measured by the fourth factor and determined that all items (i.e., 4, 7, 8, 12 and 18) pertained to experience. Originally, these SOCCS items were included in the Skills subscale; however, we determined that the presence of these items together shows promise for a fourth SOCCS subscale of Experience. The model with four subscales accounted for 54% of the variance. The Knowledge subscale was the only subscale that loaded as intended with eight items, accounting for 9.90% of variance as compared to 5.41% of variance in the original analysis (Bidell, 2005). Six SOCCS items loaded onto the Skills subscale, accounting for 27.5% of the variance as compared to 24.91% of variance in the original analysis. The remaining five SOCCS items that did not load onto the Skills subscale loaded together onto the fourth subscale, which is the Experience subscale. The Experience subscale accounted for 5.11% of the variance. Five SOCCS items loaded onto the Awareness subscale. Of the remaining items, three loaded onto both fifth and sixth factors (i.e., 11, 15, and 17). Unlike the Awareness subscale, which was theoretically justified, a fifth factor was not theoretically justified; therefore, we decided to keep these three items with the Awareness subscale.