This study examines how self-efficacy mediates students’ attitudes towards mathematics and their achievement in the subject, using the Meta-Analytic Structural Equation Modeling (MASEM) approach. After reviewing studies indexed in Google Scholar, ERIC, ProQuest, and Semantic Scholar; and assisted by AI Tools such as SciSpace and ScholarGPT, 76 primary studies met the inclusion criteria. Analyses conducted in R (via the meta package) and the MASEM web page revealed that a positive attitude towards mathematics had a moderate effect on self-efficacy ( \(\:\beta\:=0.517,\:95\%\:CI\:\left[0.446,\:0.588\right]\) ), in turn self-efficacy had a moderate effect on mathematics achievement ( \(\:\beta\:=0.240,\:95\%\:CI\:\left[0.128,\:0.352\right]\) ). Attitude towards mathematics also had moderate direct effect on mathematics achievement ( \(\:\beta\:=0.235,\:95\%\:CI\:\left[0.121,\:0.349\right]\) ), and the indirect effect via self-efficacy was statistically significant ( \(\:\beta\:=0.124,\:95\%\:CI\:[0.065,\:0.188]\) ). The model accounted for 27% of the variance in self-efficacy and 17% of the variance in achievement, leaving substantial residual variance. Furthermore, the moderator analyses that the effect were stronger at the University level with a statistically significant result ( \(\:{\chi\:}^{2}\left(3\right)=7.909,\:p=.048\) ). There were also differences in how self-efficacy was measured, especially when using the Usher and Pajares (2009) scale, which had a significant result ( \(\:{\chi\:}^{2}\left(3\right)=9.134,\:p=.028\) ). However, other moderators did not show significant differences. The results show that there are some similar but not very strong links in a wide variety of studies, and these should be seen as connections rather than proof of cause and effect. These results emphasize the importance of attitudes and self-efficacy as interrelated psychological factors that influence achievement, while also showing there are many variables across different educational environments.
