Table shows the correlation coefficients of the eight health

Table 4 shows the correlation coefficients of the eight health-related and other GBH elements in a matrix form. All the coefficients are statistically significant (p≤0.01), thereby indicating that there is at least a certain degree of relationship between the health-related and the other elements. The strongest three coefficients in each group of health-related elements are highlighted in bold.
The presence of the interrelations is determined by the magnitude of the correlation coefficients which range from weak (r=0.1–0.3), moderate (r=0.4–0.6) to strong (r=0.7–0.9) (Dancey and Reidy, 2011). By counting the number of each level of correlation, it is observed that all the coefficients demonstrate either a weak or a moderate (underlined in Table 4) correlation. Specifically, the two park quality-related dimensions, “overall quality of parks and green spaces” and “quality of facilities of parks and green spaces”, have a relatively stronger interrelation with the other non-health-related elements because 13 and ten of out 15 coefficients display a moderate level of correlation, respectively.
The results of the multiple regression analysis testing models of green brand elements are presented in Table 5. In these models, multicollinaearity is discounted for the analysis because all the variance inflation factors are below three, the critical value (Fox, 1997). Each of the eight health-related factors denoting the health components in the green brand are used as the single dependent variable separately. The other health or non-health-related elements are independent variables. The analysis of a particular health-related 17-aag included all of the other elements in the model because this maintains the integrity of a green brand. There is no ground of assumption in theory to exclude the health-related elements from the model, but the results may confirm whether these health-related elements are significant determinants in the green brand which emphasizes the other health factors. The result shows an acceptable overall fit of the models, with adjusted R-Squares ranged between 0.384 and 0.637. The models are reliable because the F-values are statistically significant (p≤0.01), whereas half of the models reach an R-Square value of about 0.6.

Discussion
The eight regression models demonstrate the underlying factors which determine the eight health-related GBH elements. Although the models have a relatively low explanatory power in predicting the influential green brand attributes against the health dimension, all the extracted elements in the models are statistically significant for drawing important conclusions about the relationships and directions between the green brand attributes which affect the health dimension of the brand (Abelson, 1995).
The values of the standardized coefficients in each model show the most significant determinants of each health-related element. It is noticeable that the key determinants of each health-related element are the other health attributes (highlighted in italics). The “pleasant experience in parks” (Model 1) is mainly determined by the perception of trees and greenery beauty (β=0.462). The “trees and greenery beauty in the city” (Model 2) is strongly determined by the perception of “landscape and hillside beauty” (β=0.438) and “pleasant experience in parks” (β=0.304). The perception of “landscape and hillside beauty” in the city (Model 3) is largely influenced by “trees and greenery beauty in the city” (β=0.529). The “recreational opportunities in parks” perceived by the respondents (Model 4) is affected by “landscape and hillside beauty” (β=0.283). The “overall quality of parks and green spaces” (Model 5) is highly determined by the health-related factors, namely, the quality of park facilities, trees and 17-aag greenery beauty and park safety (sum of β=0.829). The “quality of facilities of parks and green spaces” (Model 6) is mainly influenced by overall park quality, accessibility and safety (sum of β=0.676). The “accessibility to green spaces” (Model 7) is associated with quality of park facilities and park safety (sum of β=0.398). Lastly, “safety in parks and green spaces” (Model 8) is also determined by overall park quality, accessibility and quality of park facilities (sum of β=0.564). A significant and relatively large value of a constant and relatively lower adjusted R-Squares in this model reveals the existence of more external factors affecting the independence of recreational sites and activities in Hong Kong. These external factors may include a strong top-down urban greening policy, which causes the local residents to indulge far less in public participation and thus is probably a weaker association between health-related and other attributes in a green city brand. Its interrelation with the landscape and hillside beauty can be explained by the presence of a large coverage of country parks that are mostly located in the mountainous areas of the territory.