A data science project asking whether CDC health indicators can predict household income. Short answer: partially, and geography matters a lot.

Source Code

Role

Data Science Analyst

Timeline
January 2022

Technologies

  • Python
  • Pandas
  • Matplotlib
  • Seaborn
  • Scikit-learn

Tools

  • Jupyter Notebook

Background

Built as a team project for COGS 108 (Data Science in Practice) at UCSD, where our team investigated whether population health data can predict household income. I was curious whether conditions like diabetes or cholesterol levels correlate with income, and I focused on data cleaning, exploratory analysis, and predictive modeling using CDC public datasets.

Solution

A predictive model combining health statistics with geographic data to estimate income level. Most of the actual work went into cleaning and visualization, because the relationships only became legible once location was factored in properly.

Process

Started by surveying the CDC datasets and cleaning them. The missing and inconsistent entries were the real project. Used exploratory plots to pick features, compared modeling approaches, then wrote up findings in a report.

Final Product

Impact

Found a strong correlation between certain health indicators and economic status, with clear regional variation. A useful demonstration of what public health data can and can't tell you about economics.

Reflection

This cemented my opinion that data preparation is 80% of the job. The model was the easy part. Testing these patterns against more diverse populations would be the natural next step.