國小單步驟加減法文字題解題檢核評 量之編製發展與實測分析研究

Development and Analysis of a Checking Assessment for Word Problem Solving in Primary School Learners

本研究旨在透過成份及內容分析編製發展單步驟加減法文字題檢核評量，檢核能力成份可分成反 思及錯誤偵測二向度，反思向度由檢查行動及檢核技巧方式分層面組成，錯誤偵測向度由偵測答 案計算、運算程序、題目語意脈鉻、數量訊息擷取錯誤、正確無誤分層面組成，單步驟加減法文 字題學習內容則取自國小各版本低年級數學教科書。之後研究者使用成份及內容分析解析單步驟 加減法文字題解題中涉及之檢核能力運作與題型語意基模，接著使用所有分析結果編製發展單步 驟加減法文字題檢核評量。考量不同學校學習內容及經驗同質性，採合目標與叢集取樣抽取嘉義 縣市二年級學生864人，搭配解題經驗前導誘發評量實施單步驟加減法文字題檢核評量後，再隨 機分成兩群進行信效度交叉驗證分析。試探性及驗證性因素分析發現前述二向度各分層面因子結 構，且因素萃取、抽取變異比及模式適配度良好，並具良好效標關聯效度（以學生課堂檢核表現 教師評定分數為效標《係數及組合信度、不同向度單題、分層面及分向度極端組？考驗亦均 良好。而不同解題表現學生之解題檢核（反思及錯誤偵測）實測表現具顯著差異。

This study developed a Checking Assessment for Single-Step Word Problem-Solving for Addition and Subtraction (CA-SSWPS-AS) using componential and content analyses. The checking ability component was divided into two dimensions: reflection and error detection. The reflection dimension consisted of checking actions and checking skills， whereas error detection comprised errors in answer calculation， operational procedure， problem semantic context， and quantitative information catching， as well as a correct (no error) facet. The single-step word problems for addition and subtraction were derived from various mathematics textbooks for first and second graders in Taiwanese elementary schools. Componential and content analyses were used to examine the checking operations and semantic schema involved in single-step word problem-solving for addition and subtraction. The results were then used in development of the CA-SSWPS-AS. Considering the homogeneity of learning contents and experiences in different schools， 864 second graders from Chiayi City and Chiayi County in Taiwan were selected as participants through convenience and cluster sampling. All participants were tested using the CA-SSWPS-AS combined with experience preactivated tasks in single-step word problem-solving for addition and subtraction. The data were divided into two clusters for use in random cross-validation statistical analyses. Validity analysis using exploratory and confirmatory factor analysis produced a number of factor structures that matched the aforementioned two dimensions and their facets. The variance proportions of factor extraction and model fitness were positive， as was criteria-related validity (using students’ scores assessed by teachers in classroom performance-checking as criteria). Reliability analysis demonstrated a positive α coefficient and composite reliability. Regarding item analysis， the critical ratio was strong， regardless of whether a single item， a subassessment， or the whole was analyzed. Finally， reflection and error detection performance varied significantly when checking individuals with different word problem-solving performances.