df.understand()
Stage 3 · Day 21

Bi- & multivariate

Two columns together — check the type of both. Three combinations, three plot families. Add more variables by layering encodings onto one plot.

NUMERICAL × NUMERICAL→ scatter (relationship), then layer to go multivariate
Scatter — total bill vs tip
variables shown: 2
total_bill →tip →
layer on:

The dots already show two numbers. Each toggle encodes one more variable as colour, size, or shape — that's how a single scatter goes multivariate (up to five at once).

CATEGORICAL × CATEGORICAL→ heatmap over a crosstab
Heatmap — survival by class
DiedSurvivedClass 1Class 2Class 3801369787372119

Darker = larger. Switch to rates: raw counts say Class 3 had the most deaths, but the rate shows Class 1 passengers were far likelier to survive — the story counts alone would hide.

The real skill

The tools are the easy part. EDA is a loop: ask a question → plot it → read the story → let it raise the next question. Also in this combo family: bar/box/distplot for NUMERICAL×CATEGORICAL, plus pairplot (all numeric pairs at once) and line plots when x is time.