Chapter 3 62 after controlling for the influence of participant characteristics (age, VR knowledge, gaming frequency, prior VR experience). Preliminary analyses were conducted to ensure no violation of the assumptions of normality, linearity, multicollinearity, and homoscedasticity. When checking for multicollinearity, we observed high correlation coefficients amongst the predictor variables spatial presence, engagement, and ecological validity (>.70, see Table A5 in Appendix) indicating possible violations of the assumptions of multicollinearity. While VIF and Tolerance values were well below critical threshold, we rebuilt two iterative models omitting ecological validity in the first and engagement in the second model. Both models yielded a lower adjusted R2 (.163 and .172, respectively) than the initial model that contained all sense of presence indicators as predictors (.174); thus, we retained the initial model with all four sense of presence indicators. During the check for multivariate outliers, we have identified two cases that exceeded the maximum Mahalanobis distance value (cut-off value of 26.13 for eight predictors, see Tabachnick & Fidell, 2018, Table C4) and omitted the two cases from further analysis. Age, VR knowledge, gaming frequency, prior VR experience were entered at Step 1, explaining 3% of the variance in perceived stress (p = .460). After entry of VR spatial presence, VR engagement, and VR negative effects at Step 2 the total variance explained by the model as a whole was 23%, F (8, 123) = 4.45, p < .001. The sense of presence indicators explained an additional 20% of the variance in perceived stress after controlling for participant characteristics, R squared change =.20, F change (4,123) = 7.80, p < .001. In the final model, none of the participant characteristic variables were statistically significant. However, two of the sense of presence indicators in VR were statistically significant, with engagement recording a slightly higher semipartial correlation value (sr = .224, p = .006) than VR negative effects (sr = .218, p = .007). Therefore, only VR engagement and VR negative effects have unique contributions to the perceived stress in VR after statistically removing the overlapping effects of all other variables. Table 3.3 provides an overview of all predictor variables and the two steps of the hierarchical multiple regression model predicting perceived stress in VR. Descriptive statistics and correlation coefficients can be found in the Appendix in Tables A4 and A5, respectively.
RkJQdWJsaXNoZXIy MTk4NDMw