Phase 05 · Correlation Analysis
I ran 4 analytics notebooks against the Gold master risk profile — computing Pearson correlations, time-series trends, factor group analyses, and quintile breakdowns to identify which environmental and lifestyle factors show the strongest associations with U.S. cancer rates.
Phase Purpose
The analytics phase investigates potential associations between U.S. cancer rates and environmental, lifestyle, and food environment factors using the Gold master risk profile as the single source of truth. I compute Pearson correlations between all 17 factors and cancer outcomes, analyze 24-year incidence trends by state, and perform quintile breakdowns to test dose-response patterns.
All analysis is done at the state level (n=51). This means correlations capture macro-level patterns across states — individual-level or county-level analysis would likely reveal stronger environmental signals. The findings are presented with this caveat: statistical association does not imply causation, and lifestyle factors may mediate or confound environmental exposures.
The most important finding from the analytics phase: lifestyle factors (COPD, smoking, obesity) dominate at r=0.68 group average vs environmental factors at r=0.11 group average. However, PM2.5 exposure days emerged as a non-linear predictor surfaced only through the ML models — not visible in Pearson correlation.
4 Notebooks · Click to Explore
Analyzes 24-year cancer incidence and mortality trends (1999–2022) by state. Computes mortality-to-incidence ratios, identifies persistent high-burden states, and compares early vs late period averages to classify states as Increasing, Stable, or Decreasing.
Computes Pearson correlations between 5 AQI metrics and cancer outcomes, joins annual AQI data with cancer data for a 2000–2022 time-series comparison, and runs a quintile analysis grouping states by AQI level.
Analyzes health-based water violation trends from 1990–2024, correlates violation counts with cancer rates, and runs a quintile breakdown to test whether higher violation burden states have higher cancer mortality.
The central analytical notebook — computes a full Pearson correlation matrix between all 17 environmental, lifestyle, and food environment factors and cancer outcomes. Ranks factors by correlation strength and computes group-level averages.