Clustering Analysis Challenges Assumptions of Discrete Mathematical Skills

A study applying Bernoulli Mixture Models to exam data from 119,034 students across 13 UK national exams found that mathematical ability is largely a single, overall factor rather than a set of discrete skills. The research, published on arXiv, classified question results as pass or fail and searched for latent populations indicative of distinct skill sets. Few distinct clusters emerged, with overall student ability dominating. The best model achieved 78% accuracy, suggesting that personalized learning systems may need to reconsider assumptions about sequential skill acquisition.