"background": "Industrial machinery fleets are critical for economic productivity in developing nations, yet systematic, data-driven methodologies for evaluating their operational reliability are scarce. Existing approaches often lack the statistical rigour to account for the hierarchical nature of fleet performance data, where individual machines are nested within different sites and operational contexts. ", "purpose and objectives": "This data descriptor presents a novel methodological framework and a corresponding dataset for the multilevel regression analysis of machinery fleet reliability. The primary objective is to provide a replicable model for quantifying system reliability and identifying key explanatory variables at multiple organisational levels. ", "methodology": "A hierarchical dataset was constructed from maintenance logs, operational reports, and environmental records for a fleet of heavy equipment. System reliability was operationalised as mean time between failures (MTBF). A three-level mixed-effects regression model was specified: ijk = \0 + u{j + vk + \ Xijk +, where i denotes machines, j sites, and k manufacturers. Robust standard errors were used for inference. ", "findings": "The analysis indicates that manufacturer-level effects explain a significant proportion of reliability variation (approximately 22% of the random effect variance). Operational intensity and scheduled maintenance adherence were positively associated with MTBF, with a one-standard-deviation increase in maintenance adherence corresponding to a 15. 3% increase in MTBF (95% CI: 11. 7% to 18. 9%). ", "conclusion": "The methodological framework demonstrates that multilevel modelling is essential for accurately apportioning variance in fleet system reliability, moving beyond simplistic aggregate metrics. The accompanying dataset provides a benchmark for similar industrial analyses in comparable settings. ", "recommendations": "Practitioners should adopt hierarchical data collection practices to enable similar analyses. Future research should integrate real-time sensor data into this multilevel framework to enhance predictive maintenance models. ", "key words": "reliability engineering, multilevel modelling
Mwajuma Mwinyimkuu (Fri,) studied this question.