Skip to content
Intersectional review of correlation between food access and mortality via regression
Jupyter Notebook
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Food Deserts and the Understated Impact on Public Health


Ashley White


Many academic studies have examined the effects of food access on proximal outcomes, such as food shopping behavior and food consumption. Others have asserted the relative lack of access to full-service grocery stores and the easier access to fast and convenience foods may be linked to obesity and other diet-related diseases.

I have initiated this data science study to quantify the social externalities of food access in an effort to raise awareness and encourage communities to more effectively advocate for effective solutions.


Data Collection

Name Description Link
Food Access Research Atlas Spatial overview of food access indicators for low-income and other census tracts using measures of supermarket accessibility Link
Compressed Mortality, 1999-2016 Rates of death by underlying cause of death, state, county, age, race, sex, and year Link




Scores & Evaluation

Data Collection and Analysis

The Question:

What is the impact of food access on public healthy?

The Data Science Problem:

What is the covariance / correlation and predictive accuracy of food access features as it relates to per capita mortality rates?

The Approach:

Leverage generalized linear models and decision tree regression to evaluate mean absolute error, r^2, and feature importance


  • Food access can explain 20-80% of the variance in per capita mortality rates at the county level
  • Increased correlation in deaths from drug overdoses, transportation, and diabetes
  • Lessons to be learned from self-sustaining rural areas
  • SNAP appears to help reduce impact of food access

Next Steps

  • Incorporate Data from Food Environment Atlas & Census Tract Mortality
  • Improve viz / interactivity
  • Identify counties with biggest potential for improvement / intervention
  • Identify counties to gather lessons learned
You can’t perform that action at this time.