Impact of Item Phrasing on Detection of Differential Item Functioning: An Empirical Study of the Chinese General Self-Concept Scale

Rationale In order to avoid response bias such as agreement bias or acquiescence, it was a common recommendation that both positively-and negatively-worded items should be included in survey that the practice of using positively-and negatively-worded items may introduce systematic measurement errors that disrupt analyses and interpretations of the results (DiStefano & Motl, 2006; Quilty, Oakman, & Risko, 2006). These phenomena are referred to as method effects. So far, few related research has been conducted to examine if item phrasing affects the detection of differential item functioning (e.g., Fletcher & Hattie, 2005). This study will focus on the impact of positively-and negatively-worded items on detection of differential item functioning (DIF). Three approaches, including confirmatory factor analysis (CFA), item response theory (IRT), and the logistic regression method, are used to detect DIF items in this study. A CFA-based approach plays a key role for detecting DIF item due to its availability for incorporating method effects into the models. The major interest in this study is to explore whether DIF items exhibited in the Chinese general self-concept (CGSC) scale are affected by item wording. The more specific question is whether more negatively-worded items are tapped as DIF items. The second research question is to examine whether three different approaches used to detect DIF items produce consistent results. As far as DIF detection, we intend to investigate DIF items on the Chinese general self