TPC-Journal-14-2

The Professional Counselor | Volume 14, Issue 2 125 significant main effects in the absence of significant interaction effects. For changes in defending behavior, we used a GLM RM-ANCOVA. The independent variables and control variables paralleled the RM-MANCOVA analysis. We conducted a linear multiple regression to examine engagement of the five steps of the Bystander Intervention Model as predictors of post-training defending behavior. The five steps were entered simultaneously in the regression analysis. We calculated bivariate correlations among the criterion and predictor variables prior to conducting the main regression analyses. We examined the variance inflation factor (VIF) for predictors to assess multicollinearity. We calculated effect size for the ANCOVA models using partial eta squared (ηp 2) with .01 considered small, .06 considered medium, and .14 considered large (Cohen, 1969) and for the regression model using R2 with .01 considered small, .09 considered medium, and .25 considered large (Cohen, 1969). A p-value of < .05 indicated statistical significance. Results Preliminary Analyses Means and standard deviations for the five steps of the Bystander Intervention Model and defending behavior are presented in Tables 1 and 2. Skew and kurtosis were satisfactory and did not substantially deviate from the normal distribution for all variables. Bivariate correlations for the criterion and predictor variables are presented in Table 3. Although several of the correlations between the predictor variables were significant at p < .01, the VIF ranged between 1.08–2.69, with corresponding tolerance levels ranging from .37–.93. The VIF is well below the rule of thumb of VIF < 10 (Erford, 2015), suggesting acceptable levels of multicollinearity among the predictor variables. Changes in the Bystander Intervention Model Results of the RM-MANCOVA indicated a significant main effect for Time, Wilks’ lambda = .86, F(5, 70) = 2.32, p =.05, ηp 2 = .14., and a significant interaction effect for Time x Bystander Status, Wilks’ lambda = .77, F(5, 70) = 4.15, p =.002, ηp 2 = .23. As seen in Table 1, post-hoc RM-ANCOVAs indicated a significant main effect for Time x Know How to Act (p < .02) and Decision to Intervene (p < .01), as well as significant interaction effects for Time x Bystander Status for Notice the Event (p < .001) and Decision to Intervene (p < .05). Results indicate that Know How to Act increased from baseline (T1) to the followup assessment (T2) for both bystanders and non-bystanders. Examination of the significant Time x Bystander Status interaction effects revealed that bystanders reported an increase in Notice the Event and Decision to Intervene, whereas non-bystanders reported a decrease in engagement in these steps of the Bystander Intervention Model (see Figures 1 and 2). Changes in Defending Behavior As seen in Table 2, results of the RM-ANCOVA indicated a significant interaction effect for Time x Bystander Status for defending behavior (p < .04). As seen in Figure 3, bystanders reported an increase in defending behavior from T1 to T2, whereas non-bystanders reported a decrease in defending behavior from T1 to T2. The Relationship Between the Bystander Intervention Model and Defending Behavior As seen in Table 3, bivariate correlations revealed a positive association between post-training defending behavior and Notice the Event (p < .01), Accept Responsibility (p < .05), Know How to Act (p < .05), and Decision to Intervene (p < .01). We next conducted a linear multiple regression analysis to examine the unique effect of each of the five steps on post-training defending behavior. The full regression equation was significant, R2 = .18, F(53, 7) = 4.39, p = .002. As seen in Table 4, Notice the Event (p < .01) and Decision to Intervene (p < .05) were significant predictors of post-training defending behavior.

RkJQdWJsaXNoZXIy NDU5MTM1