Gender as a unit of analysis in clustered data: Performances of male and female learners in Gauteng and Western Cape Provinces of South Africa

Abstract

This article presents our interpretive analysis of data from the main study that was  aimed at proposing the method that will plausibly be used to analyse the Grade 12  results and to compare performances of learners between provinces of South Africa.  The participants in the study consisted of 143871 Grade 12 learners, (64208  [44.63%] male and 79663 [53.37%] female) who sat for Grade 12 examination in  2008 academic year in Gauteng and Western Cape Provinces. The null hypothesis of no differences in learners’ performances was tested. In this article we focused on testing  the hypotheses: a) female and male learners performed significantly differently (crude estimates) and b) after adjusting for province and school quintile - female and male  learners performed significantly differently (adjusted estimates). Descriptive statistics and multilevel or hierarchical models were used. Both adjusted and unadjusted models indicated that female learners performed significantly better than male learners. The methods of ANOVA (especially a balanced design) have proved to be plausible if  properly employed to clustered or correlated data.

Key Words: Gender, Unit Of Analysis, Unit Of Observation, Clustered Data, Fallacy