網路與紙本調查方式、網路問卷長短、網路問卷議題熟悉度高低的DIF分析及潛在平均數差異

DIF Analysis and Latent Mean Comparisons for Web and Paper Survey, Web Questionnaire Length, and Issue Familiarity of Web Questionnaire

本文針對過去發表的一系列網路與紙本問卷調查論文，重新進行差異試題功能（DIF）及潛在平均數的比較，分析重點亦包括問卷長短、問卷議題熟悉度對網路問卷的影響。分析過程考慮問卷內容在單向度及多向度條件下的情形。三筆資料的分析結果顯示：一、網路與紙本調查方式在電腦使用態度向度上出現DIF現象，這與過去研究一致；而在未來發展向度出現的DIF現象可能與組別間信度不一致有關；此外，信度較低的向度，可能導致潛在平均數估計不穩定而產生顯著調查方式效果；二、60題內的網路問卷長短在作答內容上並無DIF現象，但在電腦使用態度向度上有顯著潛在平均數差異，值得進一步探討；三、教育議題熟悉度高低在網路問卷某些向度上的DIF現象可能與低信度或組別間信度不一致有關，此外，並無任何潛在平均數顯著差異。本文歸納相關文獻發現，偏向個人自陳的敏感隱私、或是涉及一般時事新聞的價值判斷量表，網路問卷在作答內容的調查結果與紙本問卷間，並無實質差異存在，但在某些與電腦網路有關的問卷內容上則有部分顯著差異出現。惟上述結果的推論範圍，仍應以中小學教師母群為宜。

The aim of this paper is to analyze the past three data by detecting differential item functioning (DIF) and latent mean difference of web-based and paper-based questionnaire survey modes， and also by taking into account the influences of questionnaire length and familiarity of issue on web-based questionnaire. The analysis processes consider data under unidimension and multidimension conditions. The results show that (1) web-based and paper-based questionnaire survey modes appear DIF in computer usage attitudes dimension， which consistent with the past research. Future development dimension appearing DIF may correlate with inconsistent reliabilities between groups. Besides， lower reliabilities in some dimensions may lead to unstable latent mean estimates and significant difference between survey modes; (2) questionnaire lengths within 60 items have no DIF in the web-based questionnaire content， but have significant latent mean difference in computer usage attitudes dimension， which deserves further investigation; (3) high and low familiarity groups of educational issue in the webbased questionnaire appearing DIF in some dimensions may correlate with lower reliabilities or inconsistent reliabilities between groups. Besides， there is no significant latent mean difference existent. Finally， induction based on relevant literature finds that for questionnaires or scale involving self-reported privacy or value judgment on current affairs， surveyed outcomes in the web survey are of no substantial difference from those in mailed paper survey， but that for attitude scale toward computer and internet， some significant differences appear. Note that the external inferential effect of the above outcomes is merely appropriate for the population of elementary and secondary teachers.