Assessing measurement invariance with dichotomous items: The case of early grade mathematic assessment from the zambian sample

Abstract

Assessment of measurement invariance is an essential component of construct validity in psychological measurement. However, the procedure for assessing measurement invariance with dichotomous items partially differs from that of invariance testing with continuous items. However, many studies have focused on invariance testing with continuous items or ordered polytomous items. Hence, this study assessed the measurement invariance with dichotomous items using Confirmatory Factor Analysis. The study was conducted with data from primary school students who participated in the Early Grade Mathematics Assessment Survey (EGMA) from different provinces of Zambia in 2018. The study determined measurement invariance for three subtests of EGMA according to gender. The total sample consisted of 4698 individuals (49% male, n = 2305 and 51% female, n = 2393). The method used was a Confirmatory Factor Analysis with three highly correlated factors, number identification, number discrimination, and recognition of number patterns. The findings revealed measurement invariance of the EGMA model as defined by three subtests (number identification, number discrimination, and recognition of number patterns) across gender subgroups. However, individual analysis of each subtest revealed that the test scores have the same unit of measurement for both male and female on two factors, excluding those of number identification that lack measurement invariance. Finally, these results contribute to evidence of validity for EGMA scores in terms of their internal structure from the perspective of equality of measurement.