{ "background": "The assessment of cost-effectiveness for industrial machinery fleets in developing economies is often hampered by sparse, heterogeneous data and complex operational interdependencies. Traditional deterministic models fail to adequately quantify uncertainty, limiting robust decision-making for asset management and capital investment. ", "purpose and objectives": "This study presents a novel Bayesian hierarchical modelling framework designed to evaluate the cost-effectiveness of heavy machinery fleets. Its objective is to provide a robust, probabilistic methodology that integrates multiple data sources and explicitly accounts for operational variability and uncertainty in the Rwandan context. ", "methodology": "A three-level Bayesian hierarchical model was developed, with machinery units nested within fleet types and sites. The core model for the log-cost-effectiveness ratio of unit i is specified as () = + \ Xi + \, where \ \ (\\, \\²) represents fleet-type varying intercepts. Inference was performed using Hamiltonian Monte Carlo sampling, with model fit assessed via posterior predictive checks. ", "findings": "The model successfully quantified substantial heterogeneity in cost-effectiveness across different fleet types, with posterior distributions revealing that excavators and haul trucks were the most cost-effective asset classes. A key finding was that for haul trucks, the 95% credible interval for the cost-effectiveness ratio was 1. 4, 2. 1, indicating a significantly higher return relative to other assets. Operational downtime was identified as the most influential predictor of poor cost-effectiveness. ", "conclusion": "The proposed Bayesian hierarchical model offers a statistically rigorous framework for cost-effectiveness analysis under data constraints, providing a superior alternative to deterministic evaluations. It effectively captures uncertainty and variability inherent in industrial machinery operations. ", "recommendations": "Adoption of this probabilistic modelling approach is recommended for infrastructure agencies and private contractors to inform fleet procurement and maintenance strategies. Future work should integrate real-time sensor data to enhance model granularity and predictive capability
Uwimana Niyonsenga (Tue,) studied this question.