Leo+DadMade for Leo
Picturing Data
Rung 3 of 4 · The traps

Where Data Gets Sneaky

Data doesn't lie — but it's astonishingly easy to make it look like it's saying something it isn't. Let's meet the three classic tricks on purpose.

NESA SC4-DA1-01 Outliers · chopped axes · correlation

Explore Drag one dot far away and watch the mean lurch while the median holds. Then flick the axis toggle.
🎧
Audio WalkthroughComing Soon
Video ExplainerComing Soon

Three traps catch almost everyone. Meet them on purpose now and they'll never fool you: a stray value that drags the mean, a chart that's been quietly stretched, and a pattern that gets mistaken for a cause.

Trap One: an Outlier Drags the Mean, Not the Median

An outlier is a value sitting way out on its own — a mistake, a freak event, or something genuinely unusual. Because the mean uses every value, one outlier yanks it a long way. Take 4, 5, 5, 6 — mean and median are both 5. Now a reading comes in at 95: the mean leaps to 23, but the median is still 5. The mean is now describing a "typical" value that no actual data point is near. So whenever a dataset has a stray extreme, the median is usually the more honest middle. Drag the runaway dot in the toy and watch the mean chase it while the median holds its ground.

Say it plainly: the mean feels every value, so one wild number drags it. The median only counts positions, so it shrugs off extremes. Got an outlier? Trust the median.

Trap Two: a Chopped Axis Exaggerates Tiny Differences

Here's the favourite trick of dodgy charts. Two bars: 98 and 100. If the axis starts at zero they look almost identical — because they nearly are. But start the axis at 96 and suddenly one bar is twice the height of the other. Same two numbers, wildly different impression. Chopping the bottom off the y-axis (a truncated axis) magnifies differences that are actually trivial. It's not always cheating — sometimes you genuinely need to zoom in — but you must notice it, because it's how a 2% difference gets sold as a landslide. Flick the toggle to see honest-versus-chopped on the very same data.

Exam-saver: before you trust a bar chart, check where the y-axis starts. If it doesn't start at zero, the gaps between bars are being exaggerated — read the numbers, not the heights.

Trap Three: Correlation Isn't Causation

Two things can rise and fall together without one causing the other. Ice-cream sales and drownings both climb in summer — but ice cream doesn't drown anyone; hot weather drives both. When a chart shows two lines moving together your brain wants to shout "one caused the other!" — and most of the time you can't tell from the chart alone. There might be a hidden third thing (the heat), or it might be pure coincidence. Spotting a pattern is a great start; proving a cause needs a proper controlled experiment, not a suggestive graph.

Us, Thinking Out Loud

One reading came in at 95 when everything else was near 5. Which average would you quote, and why?

A chart shows screen time and bad sleep rising together. What would you need before you'd say one caused the other?