Rasch 模式建置國小高年級閱讀理解測驗

The Development of Reading Comprehension Test for 5th and 6th Graders Using the Rasch Model

「國小高年級閱讀理解測驗」係根據NAEP 閱讀理解評量架構，以Rasch 部份給分模式建置適用於高年級一般學童之文本閱讀理解測驗。研究對象係採分層隨機抽樣取得之高屏地區國小五、六年級學童1,052 人。測驗共39 題，含文學及訊息文本、選擇題及建構反應題，測「尋找和回憶」、「整合和解釋」及「批判和評鑑」三個閱讀理解認知層次。計量分析的結果顯示：Rasch模式建置之測驗符合客觀測量特性（閱讀理解構念為單向度、具性別及年級恆等性）、建構反應題之等級間距適切且呈階層關係、試題難度分配廣且與學童能力相對應、閱讀理解認知層次具階層性、測驗具合理之聚斂與區辨相關且能區辨不同次群體（年級、性別、語文能力）、測驗具理想之分隔信度及評分者信度。本文亦針對測驗應用、Rasch 模式應用，以及後續研究提出建議。

The Reading Comprehension Test, based on NAEP reading framework, was developed for 5th and 6th grade students using the Rasch measurement model. The participants of this study were 1,052 5th and 6th graders sampled from Kaohsiung and Pingtung counties using stratified random sampling procedures. The 39-item test, including both multiple-choice and construct-response items, measures three cognitive levels (i.e., locate/recall, integrate/interpret, and critique/evaluate) of literary and informational texts. Results show: a) the underlying trait defined by the reading comprehension test holds the characteristics of an objective measurement (i.e., the construct measured by the test is unidimensional and invariant across genders and grades); b) the rating scales for construct-response items are appropriate in terms of their order and distance; c) items spread reasonably well along the latent continuum and are aligned with various ability levels; d) the three reading cognitive levels show hierarchical structure as expected; e) the test converges and discriminates various measures appropriately; f) the test is able to discriminate group differences by gender, grade, and ability level; and g) the test has satisfactory person separation and rater reliability. Discussions and suggestions regarding test application, the application of the Rasch model, and future research are provided.