二度空間視覺化測驗之試題產生算則的驗證與修正

The Item-Generation Algorithm of Two Dimension Spatial Visualization Ability Test: Testing and Modifying

本研究以「二度空間視覺化能力」為特定的研究領域，根據林世華、劉子鍵（民86）所提出的認知測量整合模式，以及林世華、劉子鍵和梁仁楷（民86）的研究結果，進一步地探討下列問題：（一）林世華、劉子鍵和梁仁楷（民86）之研究所採用的受試皆為師大學生，使得作答反應的變異性不大，有可能因此影響估計模式的配適度。本研究進一步地加入國中二年級至高中二年級的學生作為研究樣本，使得樣本的涵蓋面擴及國中程度至大學程度，以蒐集更具代表性的反應組型，並藉此驗證模式的配適程度。（二）經由新樣本的加入，增加可能的反應組型，藉此來觀察LLTM估計結果中各成分之相對重要性的穩定程度。研究結果有數項發現異於前次研究結果。其中，針對篩選後的38題試題就整體樣本（包含大學生樣本及高國中生樣本）的資料進行Rasch模式考驗，結果發現整體樣本並不符合Rasch模式，而是符合3參數模式。此項結果與先前僅以大學生為樣本的分析結果（林世華、劉子鏈和梁仁楷，民86）大有出入。進一步針對篩選後的38題試題就高國中生樣本進行研究，卻發現該樣本資料符合Rasch模式。此一結果顯示大學生與高國中生此二組樣本在組內有相當的一致性，但在組間卻有差異性。本研究進一步針對篩選後的9題，對兩組樣本進行LLTM分析，希望瞭解兩組在各個成分上的表現是否亦有所不同。結果發現兩組樣本在基本參數之加權值的排序上有明顯的差異。此一差異將造成兩組建構出不同意涵的試題產生算則。最後，本研究基於研究結果與發現進行討論與建議。

Based on the integrated cognitive assessment model proposed by Lin and Liu (1997a) and the research findings of Lin, Liu and Liang (1997b), this study aims to investigate the following questions: (1) whether adding subjects coming from lower age levels may increase the degree of model fit of LLTM, and (2) whether the response patterns, due to the new subjects, may enhance the stability of the weights of the LLTM components. Hence, unlike Lin, Liu and Liang (1997b) whose samples were college students only, this study includes additional samples from 8th grader to 11th grader students.Compared with Lin, Liu and Liang (1997b), this study shows two significance difference. First, the data from the overall samples does not well fit the Rasch Model, in stead fits 3 pl model: the data from the high school subjects, however, fits the Rasch Model. The result suggests consistency within the same age groups but diversity between different age groups. Second, the LLTM analysis of the two age groups reveals different ranking of component weights. The difference results in constructing two difference item generation algorithms. Some discussions and suggestions are also given in the study.