多元計分人格測驗之測量恆等性：非參數方法之試題差異功能分析

Measurement Equivalence between Respondent Groups: A Non-Parametric Differential Item Functioning Analysis of Polytomous Personality Measures

自陳式人格測驗經常以李克特式多元計分試題的方式呈現。然而，此類作答方式卻容易引起對於不同應試族群是否產生了不同的測量效果之疑慮，例如，當測量目的是為進行甄選時，受試者是否可能為了獲得錄取而刻意往高分的方向填答（亦即一般所稱的「作假」），而使得測量結果和其他情境下產生差異？過去已有大量研究探討應徵者在李克特式多分題的作答是否和一般學生或在職者不同，但卻多從整份測驗的層次著手，甚少針對試題層次的測量特性進行分析。本研究運用非參數的多分題同步試題偏差檢定法（poly-SIBTEST）來進行應徵者和在職者在試題層次以及量表層次的測量恆等性分析。研究結果發現：的確有若干試題對於不同的應試族群具有差異試題功能（DIF）；然而，由於差異試題功能並無系統性地偏利於某一族群，因此在所有的五個人格量表中皆未呈現差異測驗功能（DTF）。分析結果顯示多分題人格測驗應用於甄選情境時，所測量到的潛在特性和其他情境是相等的。

The question of whether applicants respond to self-report personality measures differently when responding for selection purposes has been a crucial concern for decades. However, little research has focused on item-level measurement properties to identify the effect of testing situations on polytomous personality items. This study conducted a non-parametric poly-SIBTEST procedure to investigate both item-level and scale-level measurement equivalence on polytomous Likert-type personality scales between applicants and incumbents. The results indicated that several items exhibited differential item functioning (DIF); however, because DIF items did not systematically function with bias toward a particular group, substantial test functioning variations were not observed for all five scales. The items seemed to measure the same underlying constructs between applicants and incumbents.