Validation and application of the Chinese version of the abbreviated math anxiety scale

Evidence on the psychometric performance of the Abbreviated Math Anxiety Scale (AMAS) and the heterogeneity of math anxiety among Chinese secondary-school students remains limited. This study evaluated the Chinese version of the AMAS (AMAS-C) and examined latent profiles of math anxiety and their associations with demographic and educational variables.

Participants were 936 secondary-school students from Guangdong and Jiangxi, China. Internal consistency was assessed using Cronbach's alpha and McDonald's omega. Confirmatory factor analysis, external construct-validity analysis, and item response theory were used to evaluate the scale's measurement properties. Latent profile analysis was conducted on the nine AMAS items, followed by proportional-odds logistic regression.

The AMAS-C showed good internal consistency (α = 0.87; ω = 0.89), and its total score was negatively associated with mathematics attitudes (r = -0.57). The bifactor model provided the best fit in both the CFA and IRT comparisons. This model included a dominant general math-anxiety factor and two orthogonal context-specific factors representing residual variation associated with learning and evaluation situations. The test information function showed that the AMAS-C provided adequate measurement precision across a broad range of math-anxiety levels. LPA identified three quantitatively ordered profiles-Low Anxiety Across Both Contexts (24.6%), Intermediate Anxiety with Pronounced Evaluation Anxiety (50.0%), and High Anxiety Across Both Contexts (25.4%). These profiles differed in overall anxiety level and the relative prominence of learning and evaluation anxiety. Gender and mathematics performance were significantly associated with profile membership, whereas school stage, only-child status, and region were not. Female students had higher cumulative odds of belonging to a higher-anxiety profile than male students (OR = 1.70, 95% CI [1.32, 2.19]). Poorer mathematics performance was associated with lower predicted probabilities of first-profile membership and higher probabilities of third-profile membership. The predicted probability of second-profile membership peaked in intermediate performance categories.

The AMAS-C demonstrated satisfactory reliability and validity evidence among Chinese secondary-school students. Among the candidate models examined, the bifactor model provided a reasonable working representation of the data, with a predominant general math anxiety factor. The IRT results indicated adequate item functioning and useful measurement precision across a range of math anxiety levels. The LPA identified three ordered profiles: Low Anxiety Across Both Contexts, Intermediate Anxiety with Pronounced Evaluation Anxiety, and High Anxiety Across Both Contexts. These profiles differed in overall anxiety level and the relative prominence of learning- and evaluation-related anxiety. Female students and students reporting poorer mathematics performance were more likely to belong to higher-anxiety profiles, whereas school stage, only-child status, and region showed no statistically significant independent associations with profile membership.