Applicability of the Bayesian Sample Size Determination in Tanzania’s Demographic, Health, and Malaria Survey

Abstract

This study illustrates an exciting case of sample sizes using the Bayesian method, which rests on Bayesian decision theory and the posterior criterion. A quantitative study design was used to demonstrate (SSD) using TDHM-MIS as secondary data, consisting
of malaria prevalence among children under age five and a sample of children aged 6–59 months eligible for malaria testing. Analysis using (MCMC) simulation-based method, employed with the use of non-informative prior, the general-purpose fitting prior distribution and the informative prior as a subjective sampling prior that generates the data. The study showed that ALC has an optimal sample size, compared with ACC and WOC. Key findings suggest that the optimal sample sizes obtained were not similar, not only because of the choice of priors or lengths, but also because of the choice of SSD criteria that average over the predictive distribution of the unknown data. 95% confidence intervals favour ALC over ACC and WOC. Based on the findings, the study suggests using Bayesian techniques with highly informative priors because they reduce sample sizes to levels adequate to achieve a set of goals, as illustrated in the statistical simulation results.