4.4.2. Indirect Effects

This study estimated three mediating relationships. Table 8 shows the results of the path analysis, which tested the hypotheses of the indirect effects. The t values were calculated using the bootstrapping procedure suggested by Hayes [92], with 10,000 samples by reading the specific indirect effect from the PLS output. Table 8 shows that the t values of two indirect effects (H6 and H7) were significant at the 0.05 and 0.01 levels, respectively. Therefore, H6 and H7 were supported.


**Table 8.** Hypothesis testing (indirect effects).

\*\* *p* < 0.01, VAF (variance accounted for) = indirect effect/total effect.

Calculating the intensity of mediation is crucial for making decisions relevant to mediation effects. Following Hair et al. [87], the strength of a mediation effect was calculated by considering the variance accounted for (VAF), where VAF > 80% implies full mediation, 20% ≤ VAF ≤ 80% indicates partial mediation, and VAF < 20% does not indicate any mediation. VAF was calculated to estimate the intensity of the indirect effect by dividing the indirect effect by the total effect [93]. The VAF value indicates that approximately 75% of the total indirect effect of incivility on health is explained by the partial mediating effect of place attachment (Table 8). Hence, the relationship between incivilities and health is serially and partially mediated by place identity and place attachment, given that the VAF value is greater than 20% (i.e., 26%). Thus, indirect-only mediation was assumed, because the indirect effects were significant, but not the direct effect [87].

The R<sup>2</sup> values sugges<sup>t</sup> that incivility explained approximately 10% of the variance in place identity, whereas incivility and place identity explained approximately 14% of the variance in place attachment. However, incivility (indirectly), place identity, and place attachment reasonably explained 42% of the variance in health. Effect size (f2) was calculated to estimate the extent of the effect of an independent latent variable on the dependent variable. It is based on the change in the coefficient of determination (R2). According to Chin [94], 0.02, 0.15, and 0.35 represent a small, moderate, and substantial effect size, respectively. Table 7 shows that the f2 values for place identity and place attachment on health were 0.024 and 0.556, respectively. Thus, place attachment has a substantial effect on health.

The multicollinearity amongs<sup>t</sup> the variables in the model was also tested. The results did not emphasize any cause for concern in using variance inflation factor (VIF), the values of which were all below the suggested threshold of 5.00 (Table 7) [95]. Hair et al. [96] suggested that the predictive relevance of the model should be examined using a blindfolding procedure. The Q<sup>2</sup> values for social incivility (Q<sup>2</sup> = 0.476), physical incivility (Q<sup>2</sup> = 0.640), place identity (Q<sup>2</sup> = 0.073), place attachment (Q<sup>2</sup> = 0.100), and health (Q<sup>2</sup> = 0.356) were >0. Thus, the model had sufficient predictive relevance.
