International Journal of Academic Pedagogical Research (IJAPR)   Year: 2022 | Volume: 6 | Issue: 9 | Page No.: 36-48 Test Models and Dimensionality Assessment of Mathematics Constructed Response Items Download PDF Babatunde Kasim Oladele, Olutayo Toyin Omole, Owotorufa Ebi Obiyo Abstract: Test items from public examination bodies are expected to be valid and reliable, particularly the constructed-response items aspect of such examinations, which measures more than one trait of the test takers. The dimensionality of Mathematics constructed-response test items from two prominent Nigerian public examinations was investigated in this study. The examination bodies are the West African Examinations Council (WAEC) and the National Examinations Council (NECO). To evaluate the dimensionality of the two test forms, two relevant test models were compared. A descriptive survey design was used. The population of this study included all senior secondary school three (SSS3) students in Ibadan Metropolis in Oyo state, Nigeria. To select 24 secondary schools, a simple random sampling technique was used, and 1151 students participated in this study. For the instruments for data collection, three years of compiled past questions from the two examinations (2013-2015) were used. Classical test and item response theory models were used to analyze the data. Under the classical test theory (CTT), explanatory factor analysis, scree plot, and parallel analysis were used to determine dimensionality, whereas in the item response theory, generalized partial credit and graded response model (IRT) were used. When the CTT models were used, the results revealed that WAEC constructed-response items have four dimensions, whereas NECO constructed items have three dimensions. WAEC and NECO both have three dimensions under the IRT models. It was determined that the constructed-response test items of the two examination bodies are multidimensional and that the IRT model provided better estimates than the CTT.