Projects
A selection of projects that I'm not too ashamed of
Dynamic Flow Explorer
Go to GitHub repository
Overview
This project was developed to bring transparency and insight to the USB Alumni Association&s financial ecosystem. It visualizes the complete lifecycle of charitable donations, tracking funds from their specific sources (donor cohorts, regions) to their final destination (scholarships, infrastructure, student programs). The tool replaces static reports with an interactive experience, allowing stakeholders to explore year-over-year trends in philanthropic support.
Data
- The data consists of internal financial and donor records provided by AlumnUSB.
- Data processing involved cleaning and aggregating transaction logs to categorize funds by "Source" (Alumni Year/Group) and "Target" (Program/Allocation).
- The dataset is structured to support temporal filtering, allowing users to view cash flows for specific fiscal years.
Technologies & Methods
- Built using R and Shiny for the interactive web framework.
- Utilizes Plotly and custom D3.js integration to render the Sankey diagram, offering fine-grained control over node positioning and flow visualization.
- Implements a bilingual UI (English/Spanish) to serve the international alumni community.
Challenges & Opportunities
- One technical challenge was bridging the gap between R/Shiny and custom JavaScript to achieve specific layout requirements for the Sankey nodes that standard libraries could not handle.
- Ensuring the visualization remained readable and responsive across different devices while handling complex, multi-tiered financial flows.
Retrospective
- This project demonstrates the value of "Data Storytelling" in the non-profit sector, turning dry financial spreadsheets into an engaging narrative.
- Future iterations could include predictive modeling to forecast donation trends based on historical alumni engagement patterns.
- The integration of multimedia elements (video testimonials linked to specific data nodes) successfully humanized the data, connecting dollars to actual student impact.
Causal Impact Analysis of Policy Changes
Go to GitHub repository
Overview
This project examines the dynamics between the executive and the legislature during democratic backsliding, specifically focusing on Venezuela. It investigates how legislative constraints on the executive branch are affected during episodes of democratic breakdown and regressed autocracy. The study reveals how aspiring autocrats diminish legislative constraints and alter executive-legislative relations. It aims to determine and quantify the decline of legislative constraints and their sub-components during democratic backsliding.
Data
- This project primarily uses data from the Varieties of Democracy (V-Dem) project and the Episodes of Regime Transition (ERT) datasets.
- The V-Dem dataset provides panel data on democratization and autocratization processes, including measurements of democracy and executive-legislative relations.
- The resulting dataset includes 37 countries from 1959 to 2019.
Technologies & Methods
- This project relays mainly on the Synthetic Control Method (SCM).
- This method involves constructing a "synthetic" Venezuela, a combination of other countries that closely resembles Venezuela in political, social, and economic aspects before the onset of democratic backsliding.
- The SCM is used to compare the actual post-treatment outcomes in Venezuela with the counterfactual outcomes that would have been expected had the democratic backsliding not occurred.
Challenges & Opportunities
- One of the main limitations as the difficulty of finding countries similar to Venezuela during the period of study. The V-Dem project datasets helped overcome the issue of not having data that could be collected on the field.
- The timeframe from this project was less than six months, which diminshed the expectation of doing field work but opened the opportunity for causal inferences approaches such as SCM.
Retrospective
- With a more mixed-method approach, this project could be expanded in the future to include study cases such as Hungary in Europe.
- Another limitation and opportunity for this project should be focused on comparing the four different legislatures (2000-2006, 2006-2011, 2011-2016, and 2016-2021) during the historical period known as Bolivarian Revolution when Chávez came to power in 1998.
- I produced research results relying on computer experiments and curated dataset at a fast-paced which allowed me to comply with the time restrictions.
Network Graph Analysis for Corruption & Fraud Detection
Go to GitHub repository
Overview
It uses social network analysis (SNA) to identify key actors and the resilience of these networks. The study aims to provide knowledge on how these connections are formed and propose policy reforms to hinder their formation. The object of this project has both academic and political implications. It tests theoretical expectations of SNA in the context of corruption networks. The study assesses the international community's behavior in disrupting corruption networks, comparing it to the US's behavior.
Data
- The primary data source is the Transparency International (2020) report, "Chavismo Inc.", in collaboration with Alianza Rebelde Investiga (ARI) and CONNECTAS.
- The report covers 86 investigation cases across 61 countries, involving 751 agents (persons of interest) and 239 institutions.
- The social network includes over 3,900 relationships, including occupied functions, corruption denunciations, enablers, family ties, business connections, international trials, company connections, sanctions, contracts, friendships, student connections, enemies, company integration, human rights violations, and designations by public officials.
Technologies & Methods
- The research employs the Statnet package in R and associated libraries for network analysis.
- Techniques for assessing and disrupting dark networks are used, drawing from the sub-discipline within Social Network Analysis (SNA).
- Quadratic Assignment Procedure (QAP) and logistic regression are used to determine tie formation and the likelihood of international or US trials for corruption.
- Factor analysis is employed on centrality measurements.
Challenges & Opportunities
- Data limitations are a significant challenge when studying dark networks.
- The hidden nature of corruption networks makes it difficult to assess their organization and power distribution.
- The study seeks to fill the gap in determining the actual brokers of the network and delving into its functioning.
- Addresses the need to move beyond descriptive analysis based on observational data and centrality measurements to draw causal links.
Retrospective
- Learned that measuring prestige using cutpoint strength and brokerage level in addition to centrality offers a more precise method to determine the most important actors within the network.
- More data is needed to better characterize the actors. A more holistic approach that includes different measures.
- A potential gap to fill in future studies relates to the social dimension, given the limited data on friends, enemies, and student colleagues within the network.
