以雙因素探索性結構方程模式檢視中文版線上認知量表之因素結構

Using Bifactor Exploratory Structural Equation Modeling to Explore Factorial Structure of the Chinese Version of the Online Cognition Scale

本研究主要參考Davis、Flett與Besser（2002）的線上認知量表編修中文版線上認知量表，以驗證性因素分析、探索性結構方程模式與雙因素探索性結構方程模式分析模式適配度，並檢視中文版線上認知量表之因素結構與心理計量特質。本研究共蒐集兩份國小五年級學童資料，分兩研究進行分析，研究一資料採分層隨機叢集抽樣，抽取28所學校共2,056位學生，此研究主要目的在比較不同因素模式之適配度，並檢視量表之因素結構；研究二資料採便利取樣，將290位國小五年級學童納入分析，主要目的在交叉複核研究一之分析結果。兩份資料分析結果皆顯示一個整體因素、四個特定因素的雙因素探索性結構方程模式具有較佳且良好的適配度，中文版線上認知量表題項在整體因素的負荷量都相當高且顯著，整體因素約能解釋七成至八成的變異量，是具信、效度且定義良好的因素，可代表問題性網路使用的整體不適應認知程度；特定因素中則以社交撫慰的測量模式品質較好，可對應認知行為模式的社會孤立因素，後續研究可以此雙因素探索性結構方程模式進行分析，再以適配度良好的結構模式檢視整體認知因素或重要特定因素與其他構念間之關聯。最後依據研究結果提供在量表因素結構與模式分析上的建議。

Based on the Online Cognition Scale proposed by Davis, Flett and Besser (2002), the research modified a Chinese version of the Online Cognition Scale (COCS). The main purposes of this research were to evaluate the model fit and to explore factorial structure of COCS using bifactor exploratory structure equation modeling (BESEM), exploratory structure equation modeling and confirmatory factor analysis. Two analyses based on two data sets were conducted. The first data set was composed of 2,056 fifth graders from 28 elementary schools using a stratified cluster sampling method. The first analysis aimed to compare fit of various models and to explore psychometric properties of the COCS. The second data set was composed of a convenient sample of 290 fifth graders and were analyzed to cross-validate the main findings of the first data analysis. The results of two data analyses consistently show that a BESEM model with one general factor and four specific factors exert better model fit. The general factor may correspond to the general level of maladaptive cognition and explain up to 87% of the overall variance. It is a well-defined factor with high reliability and validity. Among the four specific factors, the dimension of social comfort has higher construct replicability and reliability. It may correspond to the social isolation factor of the cognitive-behavioral model. This BESEM measurement model may be incorporated into a general structural equation model to further explore the relationship between general maladaptive cognition factor and other constructs. Finally, suggestions for analyses of factorial structure and bifactor models are provided.