本研究嘗試建立以試題結構為基礎且適用於數學領域能力指標評量之電腦化適性測驗演算法,主要目的設定如下:: 一、開發以多點計分試題結構為基礎之適性測驗選題策略, 二、探討以多點計分試題結構及二元計分試題結構之適性測驗選題策略的成效比較, 三、提出如何結合二元計分與多點計分試題結構之適性測驗選題策略, 四、探討結合二元計分與多點計分試題結構之適性測驗選題策略的成效比較。 研究結果發現: 一、多點計分試題結構較二元計分試題結構適合用於能力指標適性測驗。 二、結合二元及多點計分試題結構之適性測驗選題策略由於個別單獨之選題策略 三、最佳指標結構之閾值與試題結構之閾值需由實徵資料中獲得
The purposes of this research are in the following. 1. Develop new polytomous item ordering coefficients for measuring the ordering relationship between polytomous items or testlets, 2. Construct adaptive testing algorithms based on the item structures estimated by the proposed polytomous item ordering coefficients, 3. Construct adaptive testing algorithms based on the hybrid item structures estimated by both dichotomous and polytomous item ordering coefficients, 4. Evaluate the performances of the proposed adaptive testing algorithms. Two polytomous item ordering coefficients are proposed for estimating the ordering structures of mathematics tests of competence indices composed by testlets. Two real data sets are used for evaluating adaptive testing algorithms based on the item structures estimated by ordering theory and two proposed coefficients, and the hybrid item structures. Experimental results show that 1. The performances of the proposed polytomous item ordering coefficients outperform that of ordering theory. 2. The hybrid item structures are better choices than others.