統整科學過程技能潛在結構分析之條件最大概度估計方法

作者:許榮富(國立臺灣師範大學物理系所)

摘要：

本研究引用近代測量理論中條件最大概度估計方法(CMLE)，針對以傳統古典測驗理論為基礎所發展並效化之TIPS(II)，進行試題及能力參數分析，除在估計方法本身之套裝程式作數位精準化修正比對外，並提出試題表徵曲線圖示法，將諸多參數作系統結構化呈現之。
本研究四項主要結論如下：一是CMLE確為分析試題潛在結構的量化模式，本文之精心分析過程及分析技術本身即為重要的測驗資料分析規範；二是TIPS(II)全卷及試題與原始之評量目標對應關係並未如古典測驗理論之分析結果；三是各試題之表徵曲線圖示法，顯示在同一難易度鑑別度下，CMLE估計法分析確實顯現高、中、低能力群之試題及能力特質；四是由信度係數及概度參數分析精準地顯示了全卷及形成假說技能之試題參數之潛在特質。本研究並建議凡涉及能力指標之相關研究諸如適性測驗等，均需以近代測量理論為理論建構，否則無法達成能力測驗之設計、分析及解釋功能。

《詳全文》

Cmle Method for Detecting Latent Structure Analysis of Integrated Process Skills

Author: Rong-Fu Hsu(Department of Physics, National Taiwan Normal University)

Abstract：

The item and ability parameters of Test of Integrated Process Skills (II) (TIPS(II)) was an-alyzed by using the CMLE based on the modern test theory in contrast to the classical test theory. In addition to revising the digital accuracy in CMLE package, the item characteristic curve (ICC) also developed in systematic structure for different kinds of variables.
Four important conclusions were drawn. First, the CMLE method is the most suitable pro-cedure for parameter estimation in 1-parameter model. Second, the test construction of TIPS (II) guided by classical test theory will not be valid for obtaining reliable measures. Third, the systemmatic ICC presenting method presented the perfect item and ability latent traits for the results in CMLE findings. Fourth, the complicated analysis of reliability coefficient and likeli-hood parameters met the precise and accurate measures of latent traits in either integrated process skills or hypothesizing skills. Also, the more intensive and applicable study in modern test theory, item response theory, latent trait theory and parameter estimation are meeded in ability test and it is recommended for further research.