Access to safe sanitation facilities remains a critical public health concern, especially in rapidly urbanizing countries like Ghana. This study investigates the determinants of household toilet ownership among property owners in three urban districts in Ghana. Using a cross-sectional survey design, data were collected from 1,256 property owners selected through a multi-stage stratified sampling procedure. Logistic regression analysis revealed that toilet ownership is significantly associated with the age and education level of property owners, community classification, building characteristics, and household income. Older property owners were more likely to own toilets (OR = 1.014 per year increase), as were those with higher education levels (OR = 1.752 for secondary, OR = 4.489 for tertiary education). Medium-class communities (OR = 2.013) completed buildings (OR = 2.625), and those constructed with sandcrete (OR = 12.755) were more likely to have toilets. Higher household income (OR = 1.00) correlated positively with toilet ownership. We conclude that enforcing building regulations requiring toilet facilities in all properties is crucial for improving sanitation in urban Ghana. Additionally, innovative sanitation financing interventions that subsidize the costs of sanitation facilities can be effective in addressing financial barriers and increasing household toilet ownership.
Stochastic models for correlated count data are crucial in reliability and safety analysis, especially when systems involve correlated events. The bivariate generalized geometric distribution (BGGD) offers a flexible framework for modeling such data. This study evaluates the influence of Beta and Kumaraswamy priors on Bayesian estimation of BGGD parameters. However, existing Bayesian studies of the BGGD have primarily relied on Beta priors without accounting for the impact of different priors. This study addresses this gap by systematically comparing the Kumaraswamy and Beta priors within a Bayesian framework and providing empirical evidence on their relative performance in estimating the BGGD parameter. Posterior distributions were derived using MCMC with Gibbs sampling, and convergence was assessed via trace and autocorrelation plots. The models were compared using Bayesian model selection criteria metrics. The findings revealed that, under the settings considered, the Kumaraswamy prior produced smaller posterior variance (0.02% to 28.51%) and better convergence diagnostics than the Beta prior (0.06% to 29.17%). The study concluded that under this investigation, the Kumaraswamy prior produced more precise posterior estimates, better convergence, than the Beta prior for correlated count processes under the BGGD. Our findings contribute to the systematic evaluation of prior distributions for the BGGD and emphasizes the practical benefits of the Kumaraswamy prior in Bayesian reliability modeling in the manufacturing industry.