Journal Design Engineering Masthead
African Structural Engineering | 08 September 2004

Replication and Multilevel Regression Analysis of Municipal Infrastructure System Reliability in Senegal

A Methodological Evaluation (2000–2026)
M, a, r, i, è, m, e, D, i, o, p, ,, A, b, d, o, u, l, a, y, e, N, d, i, a, y, e
Replication StudyInfrastructure ReliabilityMultilevel ModellingMethodological Robustness
Direct replication confirms governance capacity as significant predictor (p < 0.01).
Asset age effect 40% smaller in extended analysis, confidence interval includes zero.
Core methodological framework sound but parameter estimates show context sensitivity.
Recommends rolling-window validation for future infrastructure reliability models.

Abstract

{ "background": "Municipal infrastructure system reliability in West Africa is a critical engineering concern, yet methodological approaches for its assessment are often inconsistent. Previous studies have applied multilevel regression models, but their replicability and the robustness of their inferences in this specific context remain unverified.", "purpose and objectives": "This study aims to methodologically evaluate the replication of a seminal multilevel regression analysis for measuring infrastructure system reliability. The objective is to test the robustness of the original model's specifications and to assess the stability of its parameter estimates when applied to an expanded dataset.", "methodology": "A direct replication was conducted using the original model specification, $y{ij} = \\beta{0j} + \\beta{1}x{1ij} + ... + \\epsilon{ij}$, where $\\beta{0j} = \\gamma{00} + \\gamma{01}z{1j} + u{0j}$. The analysis was then extended using an updated, longitudinal dataset. Model performance was evaluated using robust standard errors and likelihood ratio tests.", "findings": "The replication confirmed the original finding that municipal governance capacity is a significant level-two predictor (p < 0.01). However, the effect size of asset age on failure rates was 40% smaller in the extended analysis, and its 95% confidence interval included zero, indicating statistical non-significance in the broader context.", "conclusion": "The core methodological framework is sound, but key parameter estimates are sensitive to temporal and spatial data extensions. This underscores the context-dependent nature of infrastructure reliability modelling in the region.", "recommendations": "Future studies should employ rolling-window or cross-validation techniques to test temporal stability. Engineering asset management policies should not rely on point estimates from single models without sensitivity analyses.", "key words": "replication study, infrastructure reliability, multilevel modelling, asset management, methodological robustness", "contribution statement": "This study provides the first formal replication and methodological stress-test of a key model in the region