Leo+DadMade for Leo
Analysing Datasets and Drawing Conclusions
Rung 3 of 4 · The traps

Three Ways a Conclusion Goes Wrong

The maths is the easy part. The marks — and the real-life mistakes — come from over-claiming: saying more than the data can actually back up. Here are the three traps that catch everyone.


ExplorePick a trap, judge whether the claim is fair or over-reaching, then "reveal the catch".
🎧
Audio WalkthroughComing Soon
Video ExplainerComing Soon

A conclusion should never say more than the data earns. These three traps are all flavours of the same crime — over-claiming — and examiners (and dodgy advertisers) lean on them constantly.

Trap 1 — the Sample Is Too Small

"Three kids tried our study app and averaged 8 out of 10 — everyone should use it!" Three people can't speak for everyone. With a sample that tiny, one good or bad day swings the average completely, and you've no idea whether the result would hold for the next thirty kids. A conclusion is only as strong as the amount of data behind it. Before you generalise, ask: is this enough people to be sure?

Trap 2 — One Statistic, Ripped Out of Context

"Both speakers average 5 stars, so they're equally good." Quote one number and you can hide the whole story. Two products can share a mean rating of 5 while one is dead consistent (everyone gave it 5) and the other is a lottery (ratings from 1 to 9). The averages are identical; the experiences are nothing alike. Always check the centre and the spread together — the same lesson from rung 1, now as a trap.

Say it plainly: don't over-claim. Tiny sample? Can't generalise. One stat alone? Check centre AND spread. Two things rise together? That's correlation, not proof of cause.

Trap 3 — "it Went Together, So It Caused It"

This is the sneakiest. In summer, ice-cream sales go up and sunburn cases go up. Does ice-cream cause sunburn? Of course not — hot, sunny weather drives both. When two things move together we say they're correlated, but correlation is not cause. There's often a lurking third factor doing the real work. Whenever someone says "A went up and B went up, so A caused B," ask: could something else be behind both?

The Habit That Keeps You Honest

Before you write any conclusion, run it past three quick checks: Enough data? Looked at spread, not just average? Sure it's cause, not just coincidence? If a claim trips any one of them, soften it — say what the data does show and stop there.

Us, Thinking Out Loud

How big does a sample need to be before you'd trust a conclusion?

Can you think of two things that go up together but where neither causes the other?