Heri Retnawati, Kumaidi, Gulzhaina K. Kassymova, Mam Socheath, Oscar Ndayizeye
This research study aims to find out: (1) the number of dimensions in the mathematics test, (2) the influence of the test length, (3) the number of testees, and (4) the number of dimensions to the test's relative efficiency estimated by the unidimensional and multidimensional item response theory. This simulation study used the response of testees to the mathematics test of National Examination as the data model. The focus variables in this study are the test length, the number of testees, and the number of dimensions, with 40 replications in the data generating in every case. The information function value of the data was estimated applying the item response theories, of which the results were then to be compared. The results of research indicate that (1) the test can be regarded as the unidimensional test measuring the general mathematical ability, although still measuring the specific dimensions, they are spatial ability (in the 2 dimensional model), and the spatial and numeric abilities (in the 3 dimensional model); (2) in applying the item response theory, there is a tendency that the longer the item pool is, the higher the relative efficiency to a 30-item pool is, in unidimensional, bidimensional, as well as three-dimensional models; (3) the number of testees doesn't influence the estimate of relative efficiency in the item response theory approach, (4) the more the dimension of item pool is, the higher its relative efficiency is if compared with unidimensional model. © 2022 Author(s).
Yogyakarta State University, Yogyakarta, Indonesia; Muhammadiyah University of Surakarta, Surakarta, Indonesia; Abai Kazakh National Pedagogical University and Satbayev University, Almaty, Kazakhstan; Faculty of Education, Royal University of Phnom Penh, Phnom Penh, Cambodia; Department of Languages and Social Sciences, Higher Teacher-Training School of Burundi, Ecole Normale Superieure (ENS) du Burundi, Burundi
