---
name: Exploratory Data Analysis
description: Look at the data before modeling.
---
# Exploratory Data Analysis (EDA)

EDA is the step where you form hypotheses about the data's structure before committing to a model. Skipping it means modeling assumptions you never checked.

## What to look at
- **Univariate distributions.** Histograms and box plots of every variable. Outliers and skew reveal themselves here.
- **Bivariate relationships.** Scatter plots, grouped summaries, correlation. Where do the interesting signals live?
- **Segment comparisons.** Does the relationship hold across cohorts, or does it reverse? (Simpson's paradox hides here.)
- **Missingness patterns.** Is data missing at random, or does missingness correlate with outcomes? The latter biases models.

## Output
A short list of "here's what the data suggests, here are the anomalies, here's what I'd model next". EDA is exploration, not the deliverable.