Statistical analysis

In determining associated factors affecting WTP, a linear regression model was employed. First, diagnostic analyses were performed to examine multicollinearity, heteroscedasticity, and residual normality assumptions. To assess multicollinearity assumption, the Variance Inflation Factor (VIF) was used. A VIF lower than 10 indicates there is no substantial multicollinearity between determinants in the model [44]. Heteroscedasticity and residual normality assumptions were examined using Glejser test and Kolmogorov–Smirnov test [45,46]. For both tests, a p value greater than 0.05 indicated that the residuals have a constant variance (homoscedasticity) and are distributed normally. The initial diagnostic assessments showed that the model with WTP as dependent variable did not fulfill all assumptions. Moreover, the distribution of WTP data was right skewed. The WTP data as a dependent variable were therefore converted using log transformation, which is widely accepted and have been previously [47-52]. In performing linear regression analysis, all independent variables were translated into dummy indicators where one of the categories was designated as the reference category. In the initial model, all determinants were included. Only determinants with p<0.05 in the initial model were included in the final model. The mean of the estimated WTP and its confidence interval were calculated as described previously [47,48]. The formula of Exp(Xβ^+σ^22) was used to estimate the mean of WTP, where the β^ and σ^2 were estimated regression coefficients and the mean squared error (MSE) of the regression model, respectively [53,54].