Applied Economics and Data Science: an integrated service to better understand your business and make better decisions.
News
Stochastic Climate Simulator for Agroclimatic Analysis and Risk Management
Argentina, 2026.
We developed a stochastic daily climate simulator oriented toward agroclimatic applications, designed to generate synthetic scenarios that realistically reproduce meteorological conditions relevant to agricultural production and risk management.
The model generates ensembles of daily climate trajectories that are statistically consistent with historical observations at each location, preserving seasonality, temporal persistence, intra- and interannual variability, and dependence among key variables such as precipitation, temperature, and solar radiation.
The methodology combines a stochastic model for precipitation occurrence and intensity with seasonal marginal models for temperature and radiation, coupled through a multivariate VARX model that captures the joint dynamics of the variables and their dependence conditional on wet and dry days. The simulator also incorporates a comprehensive set of diagnostics and validation metrics to assess the fidelity of simulations relative to historical records in terms of climatology, persistence, and extreme events.
This tool is particularly useful for applications such as production risk analysis, agricultural yield simulation, evaluation of management strategies under climate uncertainty, and climate impact studies in agricultural systems, where it is necessary to work with multiple plausible scenarios rather than deterministic trajectories.
Estimation of Weights for the Electricity Regulatory Index of Africa (ERI)
University of Cape Town, South Africa, 2025.
We participated in the design of an methodologically rigorous approach methodology to estimate the weights of the Electricity Regulatory Index (ERI), aimed at improving analytical robustness, cognitive efficiency in expert judgment, and empirical coherence of the index.
The methodological core of the work was the implementation of the Analytic Hierarchy Process (AHP) on graph structures using incomplete pairwise comparison matrices. Instead of requiring exhaustive comparisons, graphs were designed to ensure minimal connectivity, allowing weights to be identified from a reduced but informative set of comparisons, significantly reducing the cognitive burden on experts.
Within this framework, weights were estimated using consistent methods—such as Logarithmic Least Squares—that allow stable weights to be recovered even when comparison matrices are incomplete, while preserving traceability and internal consistency.
Additionally, Principal Component Analysis (PCA) was incorporated as an empirical tool to explore the structure of the index, identify dominant patterns and redundancies among components, and provide quantitative evidence to validate and contextualize the results derived from expert judgment.
This integrated approach successfully articulated structured expert knowledge and statistical evidence, strengthening the ERI as a regulatory evaluation tool and a support instrument for public policy formulation in the electricity sector.
Solar Panel Detection Using Satellite Imagery in Jamaica
Jamaica Public Service (JPS), Jamaica, 2025
Desarrollamos una metodología basada en el análisis de imágenes satelitales para la detección de instalaciones de paneles solares en Jamaica, orientada a mejorar la disponibilidad de información sobre generación distribuida y su integración en la red eléctrica.
We developed a methodology based on satellite image analysis to detect solar panel installations in Jamaica, aimed at improving information availability on distributed generation and its integration into the electricity grid.
The approach combines image processing techniques and spatial analysis to identify patterns consistent with photovoltaic installations, enabling the mapping of their location and territorial distribution even in contexts with incomplete or outdated administrative information.
This tool provides a strategic input for regulators and distribution companies, facilitating monitoring of distributed generation penetration, analysis of grid impacts, and the design of evidence-based policies and investments grounded in geospatial data.
Esta herramienta constituye un insumo estratégico para reguladores y empresas distribuidoras, ya que facilita el monitoreo de la penetración de generación distribuida, el análisis de impactos en la red y el diseño de políticas e inversiones basadas en evidencia geoespacial.
Analysis of Non-Technical Losses at Transformer Level for JPS
Jamaica Public Service (JPS), Jamaica, 2025
We conducted a detailed analysis of non-technical losses in Jamaica’s electricity distribution network, focusing on feeder and transformer levels, for Jamaica Public Service (JPS). The study is framed within the country’s regulatory requirements and aims to strengthen operational and analytical capacity for loss reduction.
The project combined large-scale data analysis techniques, robust regression models, and spatial analysis tools, enabling more accurate estimation of losses and differentiation of patterns by customer type, consumption level, and geographic characteristics. novel methodological approach methodologies were also incorporated to coherently assign irregular users to specific network assets.
The results provide key inputs for intervention prioritization, loss reduction strategy design, and regulatory dialogue, laying the groundwork for more efficient, targeted, and evidence-based management of the electricity distribution system.
Assignment of Illegal Users Using Network Flow Models
Jamaica Public Service (JPS), Jamaica, 2025
We designed and implemented a model to address one of the main challenges in non-technical loss analysis: the consistent assignment of illegal users to actual distribution network assets.
The problem was explicitly formulated as a bipartite network between illegal users and transformers, where feasible assignments are defined based on spatial and electrical coherence criteria. On this structure, a minimum-cost flow model was constructed, where each user represents a unit of demand and transformers act as nodes with capacity constraints.
The objective function penalizes implausible assignments—such as long distances or overloads—and identifies the globally most consistent configuration. To ensure feasibility, the model incorporates controlled discard mechanisms that explicitly identify unassignable cases without forcing artificial solutions.
This approach surpasses purely point-to-point geographic assignments, reduces structural inconsistencies, and substantially improves the quality of inputs used in subsequent statistical models. The result is a more robust, traceable, and operationally aligned estimation of non-technical losses, with direct applications in regulatory analysis and intervention prioritization.
Dynamic Sampling Methodology for Smart Meters
Jamaica Public Service (JPS), Jamaica, 2024
We developed a sampling methodology that sustains the quality of a smart meter observation panel over time, prioritizing efficient replacement when measurements are lost. The approach uses statistical criteria to select new units that compensate for missing data without oversizing fieldwork efforts.
Specifically, the procedure seeks replacements that preserve similarity between distributions (e.g., consumption) using tests such as Kolmogorov–Smirnov, and selects alternatives that meet quality thresholds with the smallest possible additional sample size, based on controlled simulation.