"background": "The management of heavy machinery fleets represents a significant capital and operational expenditure for industrial and construction sectors in developing economies. Previous studies on fleet cost-effectiveness have often relied on single-level analytical models, which may not adequately account for the hierarchical structure of fleet data. ", "purpose and objectives": "This replication study aims to methodologically evaluate the application of multilevel regression modelling for analysing the cost-effectiveness of industrial machinery fleets. It seeks to verify the robustness of this analytical approach using a contemporary dataset from the Kenyan context. ", "methodology": "A quantitative replication was conducted using operational and financial data from a fleet of earthmoving and haulage equipment. A two-level random intercepts model was specified: Cost{ij = \0j + \1X1ij +. . . + kij +, with \0j = \00 + u0j, where i indexes machinery units and j indexes project sites. Robust standard errors were calculated to account for heteroscedasticity. ", "findings": "The multilevel model successfully captured significant variance (approximately 31%) at the project site level, a factor omitted in prior single-level analyses. A key concrete result is that machinery age had a non-linear relationship with operational cost, with costs increasing by an estimated 8. 2% per annum after a threshold of seven years (95% CI: 5. 1% to 11. 3%). ", "conclusion": "The replication confirms the methodological superiority of multilevel modelling for fleet cost analysis, as it quantifies the substantial influence of contextual, site-specific factors on overall cost-effectiveness. ", "recommendations": "Fleet managers and analysts should adopt hierarchical modelling techniques to better inform capital replacement and maintenance strategies. Further research should integrate real-time sensor data into such models. ", "key words": "fleet management, multilevel modelling, cost-effectiveness,
Hassan et al. (Thu,) studied this question.