Health Data Analysis: Obesity, Nutrition, and Physical Activity

Overview

This project analyzes national and state-level health data to understand relationships between obesity rates, diet (fruit/vegetable intake), and physical activity.

The analysis includes data cleaning, feature engineering, statistical modeling, and clustering to identify patterns and key drivers of obesity across the United States.

---

Key Questions

• How does physical activity relate to obesity rates?
• Is low fruit/vegetable consumption associated with higher obesity?
• Can states be grouped into clusters based on health behaviors?

---

Key Findings

• Higher levels of physical inactivity are positively associated with higher obesity rates.
• States with lower fruit and vegetable consumption tend to have higher obesity prevalence.
• Clustering reveals distinct groups of states with similar health behavior patterns (e.g., high inactivity + high obesity).

---

Methods

Data Processing

• Filtered national and state-level health data

• Created standardized indicators:

  • obesity
  • overweight
  • fruit_lt1 (low fruit consumption)
  • veg_lt1 (low vegetable consumption)
  • pa_150 (meets physical activity guidelines)

Statistical Modeling

• Linear regression models:

  • Obesity vs physical inactivity
  • Obesity vs physical activity (≥150 min/week)
  • Multivariate regression using diet + activity variables

• Train/test split with evaluation metrics:

  • RMSE
  • MAE
  • R²

Clustering

• K-means clustering on health indicators

• Scaled variables for fair comparison

• Evaluated cluster quality using:

  • Elbow method
  • Silhouette scores

Dimensionality Reduction

• PCA used to visualize clusters in 2D space

---

Visualizations

The project includes:

• Time-series trends (obesity vs overweight)
• Boxplots of health indicators across states
• Regression plots (obesity vs activity/diet)
• Cluster visualizations (PCA projection)

---

Project Structure

scripts/
  data_processing.R     # data cleaning + feature engineering
  modeling.R            # regression, clustering, evaluation
  visualization.R       # plots

data/
  raw/                  # original dataset (not tracked)
  processed/            # cleaned data (optional)

results/
  figures/              # saved plots

---

How to Run

source("scripts/data_processing.R")
source("scripts/modeling.R")
source("scripts/visualization.R")

---

Dataset

Due to file size limitations, the dataset is not included in this repository.

Download the dataset and place it in:

data/raw/cleaned_nutrition_data.csv

---

Skills Demonstrated

• Data cleaning and transformation (R, tidyverse)
• Statistical modeling and evaluation
• Machine learning (k-means clustering)
• Data visualization (ggplot2)
• Feature engineering and pipeline structuring

---

Future Improvements

• Add additional predictors (income, education, demographics)
• Try more advanced models (random forest, regression regularization)
• Build an interactive dashboard (Shiny or web app)