Location-Dependent Bayesian Stacking for Predicting Modern Contraceptive Use Under Model Uncertainty in Nigeria

Abstract:

Introduction: Modern contraceptive use is essential for improving maternal and child health outcomes and supporting reproductive health planning. Geoadditive models provide a flexible framework for modelling modern contraceptive use by incorporating linear, nonlinear, and spatial effects. However, prediction commonly relies on selecting a single model, which ignores model uncertainty, or on conventional Bayesian Stacking with global ensemble weights, which may not adequately capture location-specific predictive performance. Aim: This study proposed and evaluated a location-dependent Bayesian stacking approach to account for model uncertainty and improve the prediction of modern contraceptive use in Nigeria. Methodology: Conventional Bayesian Stacking was modified by incorporating location-dependent ensemble weights, resulting in Geoadditive Local Stacking (GALS). The proposed approach was evaluated using data from 9,877 women in the 2024 Nigeria Demographic and Health Survey. Four competing Bernoulli Geoadditive models incorporating socioeconomic and spatial effects were fitted, and predictive performance was compared with conventional Bayesian Stacking and single-model selection using expected log predictive density (ELPD), with higher ELPD indicating better predictive performance. Results: State-level modern contraceptive use exhibited significant positive spatial autocorrelation (Moran’s I = 0.373, p < 0.001), with prevalence ranging from 3.5% in Jigawa State to 44.9% in Ekiti State. GALS achieved the highest predictive performance, with an ELPD of −0.3981, compared with −0.3983 for conventional Bayesian Stacking and −0.3984 for the best single model. Conclusion: Location-dependent Bayesian Stacking provides a reliable framework for predicting modern contraceptive use under model uncertainty and can support evidence-based family planning policies and reproductive health interventions in Nigeria.