Validation of the achievement emotions questionnaire in mathematics among high school students

Abstract

This study aims to validate the Achievement Emotions Questionnaire in Mathematics (AEQM) within a Serbian sample, comprising high school students aged 15 to 19. Grounded in control-value theory, the study posits that these emotions stem from appraisals related to control and value, subsequently influencing students' motivation, learning strategies, and performance. The AEQ-M instrument contains 60 items and assesses seven emotions (enjoyment, pride, anger, anxiety, shame, hopelessness, boredom) across three contexts (in class, during learning, in testing situations). The sample of 457 students (70% female; Mage = 16.35), from both grammar and vocational schools, completed the questionnaire either in the school or online format. Confirmatory factor analysis was carried out in Mplus 7 to examine the fit of our data to scale’s theory-proposed factor structure. A CFA of the 60 scale items showed poor model fit. After excluding 10 items with low factor loadings (< .50) and 14 items that were cross-loaded on two or more factors, the final model consisted of 36 items and resulted in adequate model fit (χ2(573) = 2194.50, CFI = .88, TLI = .87, RMSEA = .08 and SRMR = 0.07). To test reliability and convergent validity of constructs we conducted CR (> .70), AVE (> .50) and inspected factors loadings. Analysis showed that all emotions had acceptable values indicating good reliability and convergent validity. Comparing configural, metric and scalar invariance models, we tested cross-gender, school and grade measurement invariance. Based on differences in CFI < .01 and in RMSEA < .015, metric and scalar invariance was established. In the next step, we analyzed emotions’ correlations with the math grades and motivation for learning math and correlations were in expected order and directions. In order to test discriminant validity we inspected correlations between emotions and compared AVE with the squared correlation between emotions. These analyses revealed very strong correlations (> .70) between Hopelessness and factors encompassing Anxiety, Anger, and Shame, indicative of substantial multicollinearity challenges among these constructs, as obtained in previous studies. These strong intercorrelations highlight the complexity of disentangling the specific influences of these related emotional constructs, necessitating a nuanced approach in the interpretation and refinement of the analytical model.