Development of mathematics reading assessment: Psychometric evaluation based on SEM and IRT.

Abstract

The purpose of this study is to evaluate the psychometric properties of mathematics reading assessment based on structural equation model (SEM) and item response theory (IRT). The mathematics reading assessment is conducted in the form of text-reading related to real-life probability and statistics issues. The sample is 789 sixth graders from 13 primary schools in Taiwan. The latent construct of mathematics reading includes three factors which are based on the content area reading components provided by M. C. McKenna and R. D. Robinson. These three factors of mathematics reading assessment in this study are general reading comprehension, prior knowledge of mathematics, and mathematics-specific skills respectively. Firstly, the researcher develops the mathematics reading assessment and confirms the item difficulties, item discriminations, and reliability of assessment. A structural equation model (SEM) is adopted to test the factor construct of the mathematics reading assessment. According to the analysis, it shows that the item difficulties, item discriminations, and reliability are very well. The structural equation model reveals there exist three factors that conform to the assumption of content area reading components mentioned above. Secondly, this study calibrates item characteristics by the three-parameter logistic model of item response theory (IRT). Results indicate that all item fitness performs well. Items belonging to prior knowledge of mathematics have the highest item discriminations. Items of mathematics-specific skills have the highest item difficulties. This study establishes a well-structured mathematics reading assessment. Results could also provide references for instruction and assessment of mathematics reading. Finally, based on the findings of this study, some recommendations and suggestions for future researches and methodologies are also discussed.