Comparison of Bayesian dynamic linear and mixed effects interrupted time series models in assessing Uganda's neonatal mortality reduction since the SDGS onset

Abstract

The study makes a comparison of two models used to assess Uganda’s effort in reducing neonatal mortality since the onset of the Sustainable Development Goals. These two models are Bayesian Dynamic Linear and Mixed Effects Interrupted Time Series Models. The study made use of secondary data obtained from the country’s Ministry of Health spanning from January 2015 to December 2023. To determine the most appropriate model, the study conducted both Robust tests of model diagnosis and K-fold cross-validation tests as a global measure of model comparison. The Bayesian Dynamic Linear Model outperformed the Mixed Effects Interrupted Time Series Model on three out of the four criteria considered to assess the two models. Besides having a higher Cross-Validated Root Mean Square Error, the Bayesian Dynamic Linear Model does better on robust tests and has a higher Cross-Validated Log Likelihood. Therefore, the Bayesian Dynamic Linear Model outshone Mixed Effects Interrupted Time Series Model as a preferred model for Neonatal mortality analysis amidst SDGs onset impact evolution, in alignment with the study’s methodological innovations and overarching research objectives.