Page 168 - ISC PROCEEDINGS 21.4

al. (2016). The R² coefficient reflects the extent to which the predictor variables explain
the variance of the endogenous constructs in the model, with values ranging from 0 to 1;
higher values indicate stronger explanatory power. The results show that the lowest R²
value in the model is 0.077, suggesting a low level of explanatory power (Cohen, 1992).
This implies that the model explains only a small proportion of the variance in the
corresponding endogenous construct, indicating limited explanatory ability. In addition to
explanatory power, the predictive capability of the model was examined using the Q²
indicator. As suggested by Hair et al. (2016), positive Q² values indicate that the model
exhibits predictive relevance. In the present study, the lowest Q² value is 0.069,
suggesting that the model demonstrates predictive capability for the endogenous
constructs. Detailed results for R² and Q² are presented in Table 5.
Table 5. R² adjusted and Q²
R² adjusted Q²

IR 0.193 0.182

PDR 0.077 0.069

RI 0.236 0.037
Source: Authors’ own work (2026)
Multicollinearity among the predictor variables was examined using the Variance
inflation factor (VIF). The analysis revealed that VIF values ranged from 1.135 to 1.784, all
of which fall below the acceptable threshold of 5 (see Table 6). These results indicate that
the predictor latent constructs exhibit a sufficient level of independence and do not pose
serious multicollinearity concerns, in line with the recommendations of Hair et al. (2017).
Table 6. Variance inflation factor test
IR PAN PAU PDR RI

IR 1.135

PAN 1.784 1.784
PAU 1.784 1.784

PDR 1.135
RI

Source: Authors’ own work (2026)
The results of the structural model analysis indicate that most of the proposed
relationships are statistically significant. Specifically, PAN has a positive effect on IR (β =
0.263, t = 3.792, p < 0.001) and on PDR (β = 0.194, t = 3.055, p = 0.002). These findings
suggest that when virtual streamers are perceived as more human-like, users are more
likely to perceive higher levels of intrusiveness and privacy disclosure risk. Therefore, H1
and H2 are supported. Similarly, PAU has a positive impact on IR (β = 0.223, t = 3.098, p =
0.002), indicating that higher perceived autonomy increases users’ sense of intrusion
during interaction. Accordingly, H3 is supported. However, the effect of PAU on PDR is not
statistically significant (β = 0.117, t = 1.629, p = 0.103), suggesting that perceived
autonomy does not significantly increase privacy disclosure risk. Thus, H4 is rejected.
Regarding the determinants of resistance intention, both IR and PDR exert positive effects
on RI. Specifically, IR positively influences RI (β = 0.235, t = 4.501, p < 0.001), supporting

167