Analytics

HIV Policylab Policy Changes Across Region Groups Analysis

HIV Policylab policy combinations tool HIV Policylab adoption table
Python
  • Identified two sets of regions with disctinct change patterns
  • Dataset cleaning and restructuring to robustly define and carefully asses changes in policies between years
  • Challenging to handle missing data, gaps in data, changing definitions, and other complexities
  • Foundation for regression analyses I ran later in this project to asses relationship between policies and outcomes

Project Details

Company: Talus Analytics
Partner: O’Neill Institute for National and Global Health Law at Georgetown University Law Center
Partner: Georgetown University Center for Global Health Science and Security

Talus Analitics Epi Sim Javascript Implementation

Svelte
Javascript
  • Re-implementation of existing Talus epidemic model originally written in Python
  • Adapted open source interface written in Svelte to work with the Talus model
  • Model runtime changed from tens of seconds to tens of milliseconds
  • Sufficient performance for the model to display in 60fps as the user interacts with the controls, making the parameters much more intuitive
  • Allows user to manipulate 13 different paramters in addition to marking behavioral changes directly on the plot
  • User can download each model run as a CSV file
HIV Policylab policy combinations tool

HIV Policylab Policy Combinations and Adoption Table Tools

React
SVG/HTML/CSS/JS
  • Allows researchers to visualize adoption of combinations of policies
  • Options for visualizing both partial and complete adoption, terms which were challenging to define and implement correctly
  • Took on development role implementing functionality of the tool and integrating it with the existing site and datastructures
  • Worked closely with the team to iterate the precise functionality of the tool, and specific meaning of all terms
  • Implemented in pure React.js, as an early trial of what later became my DimPlot library
  • Designed and implemented single datastructure for both graph and tabular form
  • Implemented reusable filter components and system for both of these analyses and others on the site

Project Details

Company: Talus Analytics
Partner: O’Neill Institute for National and Global Health Law at Georgetown University Law Center
Partner: Georgetown University Center for Global Health Science and Security
HIV Policylab policy combinations tool HIV Policylab adoption table

Agriculture in Argentina

Python
QGIS
Illustrator
  • Long-term project demonstrating connection between Argentine transportation network and farmers' crop planting decisions
  • Created georeferenced dataset of Argentine soy crushing facilities by hand using printed data from the Bolsa De Comercio De Rosario and satellite imagery
  • Created driving dataset of duration and driving distance between every soy-producing district in Argentina and every soy-crushing facility in the country using Google Maps Distance Matrix api
  • Retrieved 7,000+ driving paths for mapping using Google Maps directions API
  • Developed econometric model for interaction between road system and planting decisions

Download Full Paper

Data Sources:

See Works Cited (too many to fit here)
Argentina Paper Main Map

Behavioral Economic Analysis of Basic Income

LaTeX
  • Economic paper summarizing most basic income studies worldwide and proposing a behavioral economic model for assessing te effectiveness of future programs
  • Principal - Agent model
  • Extensive literature review

Download Full Paper

Data Sources:

See Works Cited (too many to fit here)
Basic Income Model Quality of life and cash transfer

Long-Term Exercise & Resting Heart Rate

Python
  • Plot showing 22 months of resting heart rate (RHR) data and workout duration data leading up to and through competing in a sprint triathlon
  • Light red line indicates 30-day rolling mean RHR
  • Solid light orange indicates 30-day rolling mean workout time in minutes
  • Standard deviations are visualized in light pink to illustrate variability—a correlation between erratic workout schedules and higher RHR variability is visually apparent
  • Data scraped from Garmin Connect website using mouse automation
  • Parsed & cleaned using Python (numpy/pandas), and plotted using matplotlib

Data Sources:

RHR / Workout Minutes: Scraped from Garmin Connect webapp
Fire Risk and Powerlines Map