建構效度檢驗之線性與非線性取向：以學生創意自我效能量表為例

A Study of Examining Construct Validity on the Scale of Creative Self-efficacy for Students through both Linear and Nonlinear Approaches

在測驗編製的過程中，有時會依據兩個或兩個以上的構念來設計試題，然而為了結構上的簡潔與方便解釋的需求，研究者經常只考慮重要構念部分，而將次要構念成分歸類在未能解釋的測量誤差上。本研究旨在檢驗主要構念與次要構念在效度建立上所扮演的角色，以學生創意自我效能量表為測量工具來進行分析，研究樣本來自北部六所大學，共計636份。分析模式分別為適合連續變項分析的結構方程模式，與適合類別變項分析的線性羅吉斯測驗模式，基於此本研究除了釐清建構效度建立之合理性外，也同時驗證李克特式量表在資料分析上其尺度之本質。結果顯示在線性的驗證性因素分析中，主要構念（創意思考策略信念、創意成品信念、抗衡負面評價信念）與次要構念（獨創性、變通性、流暢性）所組成的三因素誤差相關模式整體適配度較佳。其次以非線性的線性羅吉斯測驗模式來進行分析，發現該量表符合本質上單向度的假設，且支持以主要構念組成的三因素試題特性之模式，此外以部分計分模式的參數估計值作為外在效標，也支持三因素模型的假設。

The purpose of this study is to compare two different approaches， linear structure relation (LISREL) and linear logistic test model (LLTM)， through examining constructs on the scale of students’ creative self-efficacy. The scale of students’ creative self-efficacy was initially composed of three main dimensions， but another three factors were considered as minor and ignorable. Therefore， we investigated the model-data fit for the two nested models， three-factor and six-factor models， to confirm which model is reasonable. Six hundred and thirty six subjects were sampled from the students of six universities of northern Taiwan. By linear structure relation， the results showed that six-factor students’ creative self-efficacy was better than three-factor scale. In the analysis of linear logistic test model， the six-factor model was still better than three-factor model， but the difference was much less than the difference with the method of linear structure relation. Additionally， the correlation of the estimates of difficulty parameters between the LLTM and partial credit model also showed that the LLTM was reliable. With respect to the view of measurement model rather than statistical model， three-factor model was parsimonious and efficient. The authors suggest that nonlinear model， such as item response model， is suitable to analyze the construct validity when indicator involves with main and minor dimensions.