心理學中有許多構念﹐其指標用以組合成構念﹐而非用來反映構念﹐此類構念稱之為組合性潛在構念﹐用以形成構念之指標則稱為形成性指標。組合性潛在構念在測驗編製過程中﹐多主張研究者宜清楚的定義相關面向﹐而指標需遍及各面向﹐遺漏面向可能扭曲構念意義。然對於不同研究者﹐相同構念可能有不同測量﹐構念之相關面向亦可能隨科學發展而更動﹐此即﹐在測量過程中﹐遺漏組成面向或納入非組成面向皆為可能之設定錯誤﹐本計畫即擬以蒙地卡羅法﹐探討兩類設定錯誤對模型估計值(主要是反應構念間關係之效度)與適合度之影響。其次﹐形成性指標常假設無測量誤差﹐此假設亦幾乎不可能成立﹐瞭解測量誤差之可能影響亦為本計畫研究目的之一﹐期能提供研究者在處理組合性潛在構念時之指標選擇策略。 MacKenzie, Podsakoff, and Jarvis (2005) classified latent constructs as common latent
constructs (common LCs) and composite latent constructs (composite LCs). They also
demonstrated by simulation that misspecification of the measurement model of composite
LCs results in biased estimates of path coefficients, which can invalidate the conclusion about
interrelationship between constructs. By simulation, this study shows that when
unstandardized coefficient is concerned such as in MacKenzie et al. (2005),
misspecificationresults severely inflated estimates. But if standardized coefficient is
concerned or scale is setting appropriately, the relative bias can be reduced; in some situation,
it is less than 5%.