The EDA Book

Exploratory Data Analysis in jinflow — the discipline, and the stories.

What EDA is

In 1977 John Tukey gave a name to something good analysts had always done but rarely admitted: Exploratory Data Analysis — the unglamorous, essential work of looking at the data before you model it. Not to confirm a hypothesis, but to find one. Not to prove, but to notice.

“The greatest value of a picture is when it forces us to notice what we never expected to see.” — John W. Tukey

EDA is detective work. You plot first and summarise later. You trust the median over the mean because one broken sensor shouldn’t move your conclusion — and because that broken sensor is often the whole story. You look at what the model didn’t explain, because the residual is where the surprise lives. You stay skeptical and open at the same time.

A decade later, Paul Velleman built DataDesk, and added a second half to the idea: EDA could be dynamic. Select points in one plot and watch them light up in every other. Brush a range and the whole picture responds. The data stopped being something you read and became something you could touch. (DataDesk is mostly gone now, dying quietly on the machines that still run it — part of why jinflow carries the torch.)

The principles

A short, working list — the ones this book leans on:

1. Look first. Graphics before statistics. A picture reveals what a summary hides.
2. The outlier is a clue, not noise. Resistance keeps the summary honest; the anomaly gets your attention.
3. Surprise lives in the residual — in what didn’t fit. Go there.
4. Exploration precedes confirmation. Generate the hypothesis by looking; test it afterward. Don’t skip to the trial.
5. The data is something you converse with, not just consume. (This is the half DataDesk added, and the half jinflow is built for.)
6. Honesty is a precondition. You can only trust what you see if you know what it is — its unit, its uncertainty, its provenance.

How the principles live in jinflow

jinflow didn’t set out to be an EDA tool and then discover these principles. It grew up around the same convictions, and they line up almost one-to-one:

EDA principle	In jinflow
Look first	Form follows Data — the instrument adapts to the data’s shape (a time series gets a chart, a fleet gets a map). Any perspective is a YAML declaration, not a bespoke screen.
The outlier is a clue	Severity colours the map, quality flags mark the suspect points, a silence becomes a red dot. The anomaly is surfaced, not smoothed away.
Surprise in the residual	A signal is precisely “what didn’t fit” — a balance that doesn’t balance, a reading that stops. The diamond (signal → thesis → verdict) elevates the surprise into a claim.
Exploration precedes confirmation	You brush and look to find the thread; the theses are where a hunch becomes a tested position.
Honesty is a precondition	The Typology — every reading carries its unit, its GUM uncertainty, its refinement lineage. You explore typed truth, not loose numbers. And your data stays on your machine — you can be fearless with it.

The centre of it: listen → talk

Classic EDA asks you to listen to the data — to look, and let it speak, and notice the thing you didn’t expect. jinflow adds the other direction. Because the views are linked — a click writes a selection that every panel reads (the URL is the bus) — you don’t only listen. You talk back. You point at a dot and ask “what about that one?” and the whole picture answers: the map halos the site, the chart draws its stream, the caption tells you the unit and the uncertainty.

That is the jinflow move. EDA stops being a monologue and becomes a conversation. You listen, you ask, it answers, you ask again — until the data has told you its story. Everything in this book is that dialogue in motion.

How to read this book

Each chapter after this one is a story — a real thread pulled from real data, told the way it actually happened: a thing looked different, we asked why, and the data answered. They’re not tutorials. They’re field notes from conversations with rivers, fleets, and ledgers.

• Story 01 — The Silent Optic of Coastal Bend

Written by Miss EDA. The lighthouse points; we follow. 🌊