- 期刊
「幼兒園教師助理員職務期待與職務實務評量表」之多群組測量恆等性檢定及中介效果探討
Construction and Multiple Group Confirmatory Factor Analysis of the ＂Special Education Paraprofessionals' Job Roles and Job Performances Evaluation Scale＂
摘要
本研究旨在探討「幼兒園教師助理員職務期待與職務實務評量表」之多群組測量恆等性檢定及中介效果分析。研究對象採立意取樣，邀請中部地區幼兒園中曾經在其教學的班級中實際有教師助理員進行協助的幼教師與教保員就其目前或三年內合作過的教師助理員的表現狀況填寫評量表。在研究方法與研究結果方面，首先運用結構方程模式進行驗證性因素分析，結果顯示一階三因子正交模式的適配狀況亦大致良好，只有少部分指標為尚可。第二、針對幼教師及教保進行多群組測量恆等性檢定，發現幼教師與教保員二個群組在未設限模式、因素負荷量限制模式、結構共變數限制模式、測量誤差限制模式等不同參數限制模式之間的評估指標比較上都達顯著，顯示幼教師與教保員對教師助理員職務期待的看法不同。第三，採用偏最小平方法分別對幼教師及教保員進行教師背景因素預測職務期待與職務實務的中介效果分析，發現教保員群組的中介模型的決定係數及中介效果大致良好，然幼教師群組模型的決定係數及中介效果則很低。在研究結果的解讀與應用上，本研究存在若干限制，第一，因幼兒園教師助理員幾乎都是以臨時性部分工時人員受聘，造成調查母群的不易確定。第二，調查時請幼教師與教保員針對目前或近三年合作過的教師助理員進行評量，可能有回溯或取樣上的偏差。第三，教師助理員面對的特殊幼兒並非固定，協助項目會因特殊幼兒的障礙程度及類別而存在偏差。第四，不同性質的幼兒園在聘用教師助理員的失衡狀況亦會造成取樣上的偏誤，研究者針對這些研究限制分別提出未來研究之建議，並建議未來可進行後續的質性訪談及組內分析之研究。
並列摘要
This study examines the construct validity of the ＂Special Education Paraprofessionals＂ Job Roles and Job Performances Evaluation Scale＂ through the multiple group confirmatory factor analysis (CFA). We collected the data from samples of preschools of four counties in central Taiwan, and the teachers who once worked with the special education paraprofessionals in these preschools and were willing to join our study responded to the items of our research scale. We implemented three types of analysis, firstly, we conducted the confirmative factor analysis, both the Job Roles Evaluation Subscale and Job Performances Evaluation Subscale had medium quality on most of the goodness of fit indexes. Secondly, we used Structural Equation Modeling (SEM) to perform the multiple-group analysis. We compared four types of models in different sets of cross-group constraints. The results showed significant differences between the teacher group and the educarer groups in four tested models- Unconstrained model, Measurement Weights model, Structural Covariances model, and Measurement Residuals model. We hypothesized that some background variables might cause the differences between the teacher group and the educarer group. Therefore, thirdly, we used partial least square modeling (PLS) to implement mediation models for both groups. Three formative indicators- age, diploma, and working years, estimated the teacher's background factor, as the exogenous variable. For educarer group, the medium coefficient of determination and total direct and indirect effect showed the model was just good enough. However, the teacher group had low explained variance and low total effect. We discussed these results and proposed some suggestions for the future study.