Learner Performance in the Mpumalanga and the Eastern Cape Provinces, South Africa : Why Disaggregated Analysis?

This cross-sectional quantitative study compared the likelihood of learners passing Grade 12 between the Mpumalanga and Eastern Cape provinces. The secondary clustered data used comprised of 114775 (60748 [52.93%] Eastern Cape and 54327 [47.07%] Mpumalanga) learners who sat for examinations in 2008. This study demonstrated how the ordinary and the hierarchical logistic regression models handle clustered observations for binary study end-point. The results showed that the ordinary logistic regression model produced misleading, smaller standard errors, due to inflated sample size, and falsely narrower confidence intervals. The disaggregated approach's crude and adjusted models indicated that leaners in Mpumalanga were significantly more likely to pass Grade 12 than learners in the Eastern Cape. Therefore, Mpumalanga provided better conditions for learners to pass Grade 12 than did the Eastern Cape. This paper argues that the analysis of clustered data is best achieved by the employment of statistical techniques that account for correlation between observations within clusters.