Journal Design Engineering Masthead
African Civil Engineering Journal | 07 January 2005

Replication and Validation of a Time-Series Forecasting Model for Efficiency Gains in Rwanda's Power-Distribution Infrastructure (2000–2026)

S, a, m, u, e, l, H, a, b, i, m, a, n, a, ,, M, a, r, i, e, C, l, a, i, r, e, U, w, a, s, e, ,, C, h, a, n, t, a, l, M, u, r, e, k, a, t, e, t, e, ,, J, e, a, n, d, e, D, i, e, u, U, w, i, m, a, n, a
DOI: https://doi.org/10.5281/zenodo.18971069
Replication StudyARIMA ModellingInfrastructure EfficiencyForecast Validation
Replication confirms core ARIMA model structure but identifies an 8.5% forecast overestimation.
Parameter confidence intervals are wider than originally reported, indicating greater dynamic uncertainty.
Findings underscore the necessity of periodic model re-calibration with new operational data.
Suggests integrating exogenous variables like climate and maintenance cycles in future specifications.

Abstract

{ "background": "Time-series forecasting models are critical for planning and investment in power-distribution infrastructure. A previously published model for predicting efficiency gains in a national grid has been influential but has not been independently validated in a different operational context, raising questions about its generalisability.", "purpose and objectives": "This study aimed to replicate and critically evaluate the methodological robustness of the published forecasting model. The objective was to assess its predictive accuracy and parameter stability when applied to an updated, extended dataset from the same national context.", "methodology": "This replication study applied the original autoregressive integrated moving average (ARIMA) model, specified as $\\Delta yt = \\phi1 \\Delta y{t-1} + \\theta1 \\epsilon{t-1} + \\epsilont$, to an extended, verified dataset. Model performance was evaluated using root mean square error (RMSE) and mean absolute percentage error (MAPE). Parameter uncertainty was assessed using 95% confidence intervals derived from robust standard errors.", "findings": "The replication confirmed the model's core structure but revealed a systematic overestimation of efficiency gains by approximately 8.5% in out-of-sample forecasts. The confidence interval for the primary autoregressive parameter $\\phi_1$ was found to be wider than originally reported, indicating greater uncertainty in the model's dynamics.", "conclusion": "While the original model's theoretical framework is sound, its predictive precision for long-term planning is less certain than initially suggested. The findings underscore the necessity of periodic model re-calibration with new data.", "recommendations": "Infrastructure planners should incorporate the identified forecast bias as an adjustment factor. Future research should integrate exogenous variables, such as climate data and maintenance investment cycles, to improve model specification and forecast accuracy.", "key words": "replication study, time-series forecasting, power distribution, infrastructure efficiency, ARIMA modelling, validation", "contribution statement": "This study provides the first independent validation and methodological critique of a