遊客滿意度驗證性因素分析及其模式修正之研究－－以惠蓀林場森林遊樂區為例

Confirmatory Factor Analysis of Travel Satisfaction and Model Modification Case Study in Huisun Forest Recreation Area

本研究透過遊客問卷方式，調查惠蓀林場森林遊樂區遊客滿意度之態度資料，藉由結構方程模型之驗證性分析方法，闡明遊客滿意度之潛在影響因素與測量變項間之關係。LAMBDA-X參數估計值結果顯示，各測量變項分別與其潛在變項因素之模式關係是成立的而且達顯著水準，以天然資源環境因素而言，遊樂區資源獨特性與森林景觀美質及野生動物豐富程度最能表示其特性；經營管理因素而言，設施維護管理、步道維護管理及環境清潔維護最能顯示因素特性；公共及遊憩設施因素而言，則以旅遊安全設施、遊樂設施及解說設施最能反應其適切性。PHI參數估計值顯示，僅天然資源環境因素與收費價格因素間未達顯著水準，其餘均達顯著水準，即其餘各因素間均有顯著關係。THETA-DELTA參數估計值顯示測量誤差達顯著水準，其中測量殘差估計值較大者為收費站服務品質（0.48）、野生動物豐富度（0.45）、公共廁所滿意度（0.44）、行銷旅遊資訊（0.43）及聯外道路交通便利性（0.41），測量殘差大則表示變項被潛在變項解釋的程度低。多元相關平方數值反映了測量變項能夠被潛在變項解釋的百分比，也反映了測量變項的信度（Reliability），最高者為資源獨特性變項，有68%的變異量能夠被潛在變項解釋，其次為設施維護管理變項佔64%，步道維護管理變項佔63%、森林植群環境變項62%及旅遊安全設施變項62%。經模式配適度檢定結果顯示假設模型與觀察值間有顯著差異，即理論模式無法配適實際資料，模型仍有改善修正空間。經利用修正指數等方法作模型修正，結果顯示WLSχ2值明顯下降，有明顯改善模型契合度；平方概似平方誤根係數RMSEA亦顯示模型契合度良好，其中依全部最大修正指數MI之結果最小。整体而言，依全部最大修正指數MI及刪除因素負荷量0.5以下之模式修正結果較佳。 This study attempts to analyze the relationships between visitor’s travel satisfaction andlatent variables by using confirmatory factor analysis in the Hui-Sun forest recreation area. The estimatedparameter of Lambda-X showed the model’s relationship come into existence and significant that betweenobserved variables and latent factors. The uniqueness, scenic beauty and wildlife’s abundance can mostlyindicate the characteristics of natural resource factor. The facilities maintenance, path maintenance andenvironment cleaning disclose the characteristics for management factor. As for public and recreationalfacilities, those for visitor’s security, recreation and interpretation are the most important factors. On theother hand, the estimated parameter of Phi reveals the relationship among the latent factors is significantleaving out those between natural resource and price factors, and the estimated parameter of Theta-Deltaindicates all the measurement errors are significant. In addition the magnitude of measurement errorsimplied the degree of interpretation by latent variables in negative direction and the rank in order areservice of charge station (0.48), abundance of wildlife (0.45), public restroom (0.44), marketing andtravel information (0.43), accessibility in traffic (0.41). Finally, the squared multiple correlations for Xvariablereflected not only the percent of observed variables which could be interpreted by latent ones, butalso that of their reliability. Ranking the observed variables uniqueness (68%) holds the highestpercentage, then facilities maintenance (64%), path maintenance (63%), forestry vegetation (62%) andvisitor’s security equipments (62%).The primary results show that there is significant difference betweeninitiated model and observed variables, that is, the hypothesized model can not fit with actual data verywell. Consequently, model modification is executed according to the modification indices. The outcomerepresents that WLSχ2 value decreased which can improve the model fitting. The RMSEA valuerepresents the same as well. In short, the results are much better when try to modify the model, beginningwith maximum modification indices and deleted the factors loading below 0.5.

This study attempts to analyze the relationships between visitor’s travel satisfaction andlatent variables by using confirmatory factor analysis in the Hui-Sun forest recreation area. The estimatedparameter of Lambda-X showed the model’s relationship come into existence and significant that betweenobserved variables and latent factors. The uniqueness, scenic beauty and wildlife’s abundance can mostlyindicate the characteristics of natural resource factor. The facilities maintenance, path maintenance andenvironment cleaning disclose the characteristics for management factor. As for public and recreationalfacilities, those for visitor’s security, recreation and interpretation are the most important factors. On theother hand, the estimated parameter of Phi reveals the relationship among the latent factors is significantleaving out those between natural resource and price factors, and the estimated parameter of Theta-Deltaindicates all the measurement errors are significant. In addition the magnitude of measurement errorsimplied the degree of interpretation by latent variables in negative direction and the rank in order areservice of charge station (0.48), abundance of wildlife (0.45), public restroom (0.44), marketing andtravel information (0.43), accessibility in traffic (0.41). Finally, the squared multiple correlations for Xvariablereflected not only the percent of observed variables which could be interpreted by latent ones, butalso that of their reliability. Ranking the observed variables uniqueness (68%) holds the highestpercentage, then facilities maintenance (64%), path maintenance (63%), forestry vegetation (62%) andvisitor’s security equipments (62%).The primary results show that there is significant difference betweeninitiated model and observed variables, that is, the hypothesized model can not fit with actual data verywell. Consequently, model modification is executed according to the modification indices. The outcomerepresents that WLSχ2 value decreased which can improve the model fitting. The RMSEA valuerepresents the same as well. In short, the results are much better when try to modify the model, beginningwith maximum modification indices and deleted the factors loading below 0.5.