國家教育研究院

　　認知診斷模式一般用於單元測驗，透過分析受試者之認知反應組型，瞭解學生的學習狀況，包括認知概念或技能精熟程度，作為現場教師補救教學分組參考。本研究嚐試利用較大內容範圍之縣市五年級數學學力檢測資料進行分析，目的是以認知診斷模式分析縣市學力檢測資料，探究試題的性質，如鑑別度、猜測參數、疏忽參數與認知概念或技能精熟程度等，並進行模式與試題適配度探討，以了解認知診斷用於成就測驗或學力檢測或現場教師補救教學分組參考之可能性。

　　分析結果發現，學力檢測試題具有不錯之信效度。因軟體限制，認知診斷模式只能以抽樣及大範圍認知概念或技能之Q 矩陣分析，在模式適配性方面，DINA 與GDINA 雖有部分指標未能符合標準，但大體而言在模式與試題上皆有不錯之適配度。在四個領域中認知概念或技能屬性平均以數與計算、代數、統計最高，而以圖形與空間最低；在所有認知概念或技能屬性中，以會讀寫做十萬以內的數精熟度最高，而以圖形與空間之了解三角形的兩邊和大於第三邊或兩邊差小於第三邊精熟程度最低─未達一半。另於本研究亦發現屬性精熟度與大多數對應試題之答對率接近，DINA 猜測參數與粗心參數估計和高低分組之答對率具有高度相關。最後建議認知診斷模式在軟體未臻完善前，仍考慮以較小範圍內容之測驗與人數2000 以下為宜，以便所得結果可以運用於後續補救教學，提升學生學習成效。

　　Cognitive diagnostic model (CDM) is generally used for unit tests, through the analysis of the examinee's cognitive response patterns to understand student learning results, including cognitive concepts or skills mastery level. Teachers can use it as a field reference to group students and do remedial instruction. This study attempts to use a greater range of content of the fifth grade math competency test data to analyze in one county. The purpose is to analyze the county’s competency test data by cognitive diagnostic model in order to explore the items’ properties, such as discrimination, guessing parameters, slip parameters and cognitive concepts or skill mastery level as well as model’s goodness of fit. In order to understand the possibility that cognitive diagnostic test is applied for achievement tests, competency test or remedial instruction grouping reference.

　　The results found that the competency test items with a good reliability and validity. Due to software limitations, cognitive diagnostic model can only analyze a part samples from population and Q matrix with a wide range of cognitive skills or concepts. About model goodness of fit, although some indicators of DINA and GDINA models do not correspond to the criteria, but generally speaking on the model and items of the goodness of fit are acceptable. Among the four areas, the area of number and calculation, algebra, as well as statistics has the highest mastery of cognitive concept or skill attributes. And the area of graphic and space is the lowest. Among the all cognitive concepts or skill attributes, “can read, write, and calculate the number of within one hundred thousand” has the highest mastery. And “the sum of two sides of a triangle is greater than the third side or the difference of two sides of a triangle is less than the third side” has the lowest mastery which smaller than .5. In addition, this study also found that attribute mastery and answer correct rate correspond to the same item with high correlation. The guessing and slip parameters of DINA model have high correlation with the answer correct rate of high and low score groups. Finally, the study recommends that the test has small context and the sample less than 2000 will be appropriate before the software of cognitive diagnostic model might not be perfect. So that the results could be applied to subsequent remedial instruction to enhance students' learning.