頁籤選單縮合
題 名 | 大型測驗中同時進行垂直與水平等化效果之探討=A Simultaneous Vertical and Horizontal Equating of Large-scale Assessments |
---|---|
作 者 | 郭伯臣; 王暄博; | 書刊名 | 教育研究與發展期刊 |
卷 期 | 4:4 2008.12[民97.12] |
頁 次 | 頁87-119 |
分類號 | 521.3 |
關鍵詞 | 平衡不完全區塊設計; 定錨不等組設計; 定錨試題; 測驗等化; Balanced incomplete block design; Non-equivalent groups with anchor test design; Anchor item; And test equating; |
語 文 | 中文(Chinese) |
中文摘要 | 近年來,隨著資訊科技快速進步、測驗形式的改變及需求量的快速增加,大型測驗(large-scale assessments)的議題廣泛受到矚目。然而這些大型測驗主要的目的為建置一套完整且客觀的學生學習成就資料庫,並透過等化連結的方法使不同年級、不同年度的受試學生測驗分數可進行比較,進而了解全國學生之學習成效。本研究以試題反應理論(item response theory, IRT)之三參數羅吉斯模式(three-parameter logistic model)為理論基礎探討利用平衡不完全區塊(balanced incomplete block, BIB)設計與定錨不等組設計(non-equivalent groups with anchor test design, NEAT)兩種連結設計在進行大型教育測驗等化時,對於不同年級不同年度間等化之連結效果,並針對受試人數、定錨試題比例及難度範圍選取法等變項進行模擬實驗。在本研究發現試題參數及受試者能力值估計誤差隨著受試人數增加而減少;以定錨比例來看,隨著定錨比例的增加,受試者能力估計誤差及試題難度參數估計誤差減少;以選題範圍來看,難度範圍並沒有明顯差異;BIB設計於試題參數估計精準度大致上優於NEAT設計;NEAT設計受試者能力估計精準度較優於BIB設計。 |
英文摘要 | For large-scale assessments, the spectrum of subject matter is usually wide and the simultaneous sampling of items and students is a practical way to obtain representative indications of student performance. Balanced incomplete block (BIB) design and non-equivalent groups with anther test design (NEAT) are two popular test equating methods for this condition. The purpose of this study is to explore the linking performances of two large-scale assessments which are administrated in different years and different grade by using BIB and NEAT designs. The effects of numbers of people, the percentage of anchor items and ranges of anchor items are explored under two different equating designs.The results of simulation study show that: 1. the estimation error decreases as the numbers of people increases; 2. the estimation error decreases as the number of anchor items increases, and the better equating performance occurs as the percentage of anchor items is 30%; 3. BIB outperforms NEAT in estimating item parameter and NEAT outperforms BIB in estimating abilities of examinees for two different equating designs. |
本系統中英文摘要資訊取自各篇刊載內容。