Measurement Invariance in the Graduate Record Examination: A Language-Based Analysis

Abstract

Objectives: The study aimed to examine the measurement invariance in the GRE test for quantitative reasoning administered in both Arabic and English. Method: A cluster random sample of 541 male and female MA students was selected from the University of Jordan, Yarmouk University, and Mutah University. Of these, 273 students (both male and female) took the test in Arabic, while 268 students (both male and female) took it in English. The data were analyzed using both confirmatory factor analysis (CFA) and multi-group conformity factor analysis (MGCFA) through the statistical program AMOS 24. The fit of the data to the factorial model was assessed using four indicators: χ²/df, CFI, RMSEA, SRMR. Results: The results indicated that the data fit the four-factor model (arithmetic, algebra, geometry, and data analysis) for both the Arabic and English versions of the exam. However, the results also showed that while configural invariance was achieved, metric and scalar invariances were not achieved based on the language of the exam. Conclusions: The quantitative reasoning exam presented in both English and Arabic versions had the same structure as assumed in the factorial model. However, the loadings on the factors and the intersection between the factors differed between the two versions. Preference was given to the Arabic version of the exam because it demonstrated a good fit with the factorial structure model, and proficiency in the English language did not significantly influence performance. Therefore, using the Arabic version with students whose mother tongue is Arabic allows for fair and unbiased comparisons between them.