Neutrosophic and Multinomial Analyses of How Foundational Skills Gatekeep Mathematics Cognitive Depth in Ghanaian Tertiary Education
Eunice Osei-Asibey
*
Department of Mathematics and Computer Studies, Ada College of Education, P.O. Box AF34, Ada – Foah, Ghana.
*Author to whom correspondence should be addressed.
Abstract
This study examines how domain-specific mathematics competencies predict students’ Depth of Knowledge (DoK) classifications in a Ghanaian tertiary institution using a multinomial–neutrosophic analytical framework.
The study adopted a cross-sectional, single-institution design involving a census of 51 eligible students at Ada College of Education, Ghana. Multinomial logistic regression was used to examine associations between mathematics competencies and student-level Depth of Knowledge (DoK) classifications. Separate composite-score and domain-specific specifications were considered, while mediation, moderation, and neutrosophic analyses were used to explore conditional relationships and uncertainty in the estimated associations. Bootstrap resampling and receiver operating characteristic analyses provided supplementary assessments of coefficient stability and model discrimination. Given the institutional scope and modest sample, the findings are interpreted as exploratory, cohort-specific associations rather than population-level estimates applicable to Ghanaian tertiary education generally.
Data were obtained from 51 students at Ada College of Education and included scores in Trigonometry (MTRIG), Functions (MFUNC), Algebra (MALG), Probability (MPROB), and Calculus (MCALC), together with a composite Total Mathematics Score (TMAT) and DoK Levels 1–4. Multinomial logistic regression was used to model direct domain effects on DoK classification, while mediation and moderation analyses assessed the role of foundational mathematics (M1). Neutrosophic analysis was employed to quantify truth, indeterminacy, and falsity in the estimated relationships, and model discrimination was evaluated using receiver operating characteristic (ROC) curves. Results indicate that TMAT is a significant positive predictor of DoK Level 4 (p = 0.0022), whereas MPROB exhibits significant negative associations with DoK Levels 3 and 4. Mediation analysis shows that TMAT exerts consistently negative indirect effects on all DoK levels through M1, with the strongest suppression observed for DoK Level 4. Moderation analysis reveals that M1 significantly appears to strengthen the effects of TMAT and selected domain scores on cognitive depth, particularly for higher DoK levels. ROC analysis demonstrates strong classification performance across all DoK categories (AUC > 0.93). An alternative explanation is that probability assessment frequently emphasizes abstract reasoning and uncertainty rather than procedural accumulation of mathematical skills. Consequently, students who perform well in probability may not necessarily demonstrate uniformly high performance across all domains contributing to the composite cognitive-depth classification. The observed negative coefficient should therefore be interpreted as a conditional association within the fitted multinomial model rather than evidence that probability knowledge impedes cognitive development. These findings highlight the conditional and uncertainty-aware nature of cognitive depth progression in mathematics and underscore the importance of integrating foundational skill reinforcement with advanced domain instruction in tertiary mathematics education.
Keywords: Cognitive development, depth of knowledge, foundational mathematics, multinomial regression, neutrosophic analysis