The Number You Computed Is Not the Number You Were Asking About

You'll learnto tell the group you measured apart from the group you were asking about — which number is the statistic, which is the parameter, why they differ without anyone making a mistake, and why a bigger sample from the wrong place never closes the gap.

A city has 200 registered voters and a real answer to the question "what share of them support the mayor?" — and nobody can see it. A poll can afford to ask 25. This animation builds that city, draws that sample, counts what comes back, and then cheats once and measures everyone, so you can watch the poll's 52 percent sit eight points away from the city's 44 with nothing done wrong. Then it polls the city forty more times, and finishes on the repair that fails: asking more and more people from one neighbourhood, which converges beautifully on the wrong group's answer. Twelve short steps; press Next for each one.

12 STEPS · 6 QUICK CHECKS · POPULATION vs SAMPLE · PARAMETER vs STATISTIC · SAMPLING VARIATION · BIAS · v2

The city's 44% never moves; the poll's 52% belongs to another group.

Before you start

What you're looking at

200 circles, one per registered voter in a small city. A circle stays hollow until someone asks that voter; only then does it show a color.

The question

What share of all 200 support the mayor — and what has a poll of 25 of them actually told you about it?

Step 1 / 12