72 The Professional Counselor | Volume 12, Issue 1 as a whole), (e) homoscedasticity, (f) absence of multicollinearity, (g) no significant outliers, and (h) normally distributed residuals (Flatt & Jacobs, 2019). The research variables met assumptions (a) and (b) in conducting multiple regressions. In analyzing data in SPSS, independence of residuals was determined by using the Durbin-Watson statistic, which ranges in value from 0 to 4, with a value near 2 indicating no correlation between residuals. Assumption (c) was met, as the Durbin-Watson value found was 1.46 (Savin & White, 1977). Additionally, we plotted a scatterplot using variables, as well as a partial regression with each of the independent variables and the dependent variable, and observed linear relationships, attending to the assumptions of linearity (d; i.e., a linear relationship between dependent and independent variables) and homoscedasticity (e; i.e., residuals are equal for all values of the predicted dependent variable). Homoscedasticity was also assessed by visual inspection of a plot of studentized residuals versus unstandardized predicted values. To assess the absence of multicollinearity (f), we considered the variance inflation factors (VIF) indicated in the coefficients table (Flatt & Jacobs, 2019). We found VIF values ranging from 1.01 to 1.05, indicating an absence of multicollinearity (f). VIF is a measure of the amount of multicollinearity in a set of multiple regression variables (Flatt & Jacobs, 2019). We checked for unusual points (g): outliers, high leverage points, and highly influential points. We did identify a significant outlier (−3.10) in case number 133 by examining the range of standardized residuals ([−3.10 to 2.34]), which is outside the common cut-off range of three standard deviations (SD). We then inspected the studentized deleted residual values and found a value in case number 133 (−3.15), which falls outside the common cut-off range of 3 SD. Additionally, we determined two cases of problematic leverage values that were greater than the safe value of 0.2 (0.36 and 0.23). The cases that violated assumptions were filtered out and the standard multiple regression analysis was run again. This time, the data did not violate assumptions (a) through (g). Last, we observed normally distributed standardized residuals (h). To determine if any cases were influential in the data, we examined the Cook’s Distance values, which ranged from .000 to .090. As there were no values above 1, there were no highly influential points. To answer the first research question (the relationship between PSC factors and levels of PSC self-perceived competence in working with trans students in schools as measured by total scores on the GICCS), we used a standard multiple regression analysis (Sheperis et al., 2017). To answer research questions 2 through 4, we conducted standard multiple regression analyses using the Awareness, Knowledge, and Skills subscales as the dependent variables, respectively. Results Correlations Between Variables of Interest Prior to the regression analysis, we examined correlations between the variables: PSC GI (cisfemale, cismale, trans, agender), having worked with trans students, postgraduate training experiences, personally knowing someone who is trans, the GICCS Awareness subscale, the GICCS Skills subscale, the GICCS Knowledge subscale, and the GICCS total score. Correlations of variables of interest are found in Table 2. There were multiple significant correlations as determined by Pearson product moment correlations (r). The GICCS total score was significantly correlated with the Awareness subscale (r = −.65, p < .001), the Skills subscale (r = .83, p < .001), and the Knowledge subscale (r = .66, p < .001). The Awareness subscale was significantly correlated with the Skills subscale (r = −.26, p < .001) and the Knowledge subscale (r = .30, p < .001). The Knowledge subscale was also significantly correlated with the Skills subscale (r = .30, p < .001). In examining demographic factors, cisfemale GI was significantly correlated with cismale GI (r = −.90, p < .001), trans GI (r = −.37, p < .001), and agender GI (r = −.21, p < .001). Additionally, cisfemale GI was significantly correlated with having worked with trans students (r = −.12, p = .036), as well as the GICCS total score (r = −.14,
RkJQdWJsaXNoZXIy NDU5MTM1