Bayesian Hierarchical Spatial Modeling and Mapping of Adult Illiteracy in Kenya

Abstract

Regional disparity in literacy levels must be addressed if Kenya is to achieve its international goals such as Education for All (EFA) and Millennium Development Goals (MDG). Literacy level in Kenya has been on the rise. However, the 2007 Kenya National Literacy Survey crude rates showed that on average 38.5 per cent (7.8 million) of Kenya’s adult population was illiterate with significant regional and gender variation.

Bayesian binary logistic models (with and without CAR spatial and unstructured random effects)  are applied to the Kenya National Adult Literacy Survey (2007) data that was obtained from sampled 18000 households, 4782 in urban and 10914 in rural areas, to investigate spatial variation of illiteracy levels in Kenya. There were 15734 successful interviews that were comprised of 6493 were male and 9241 female

The best fitted model was found to be the CAR model with age, sex, disability and awareness of adult literacy programs as the significant explanatory variables. Smoothed map of illiteracy from the best fitted model was then produced together with its corresponding confidence interval maps for regional variation in Kenya, in order to capture visual uncertainty in estimation. These maps can be used by policy makers to identify the pattern and tailor make programs appropriate for each region.

Keywords: Illiteracy, Bayesian Hierarchical Models, Spatial modeling