Two Randomnesses, and Neither One Does the Other's Job

You'll learnwhich conclusion each kind of randomness buys you, by running both on the same 240 students and measuring what each one actually fixed.

A school of 240 students, 40% of whom already study regularly. Draw 60 at random and you get 41.7% — the school, near enough. Draw 60 from the 8 a.m. library and you get 80%, which is a fact about the library. That is what random selection buys: a sample that stands for the population. Now take the random 60 and split them for a study-skills workshop. Let them choose and the workshop group holds 20 regular studiers against 5; toss a coin and it holds 13 against 12. That is what random assignment buys: two groups alike in everything you did not measure. The workshop in this file has no effect whatsoever — no score depends on which group a student is in — and the self-selected comparison still reports it 6.95 points ahead. Twelve short steps; press Next for each one.

12 STEPS · 6 QUICK CHECKS · THE QUESTION · FOUR COLLECTION METHODS · SELECTION · ASSIGNMENT · CONFOUNDING · SCOPE · v2

selection buys reach · assignment buys cause · the workshop effect here is 0

Before you start

What you're looking at

240 circles, one per student at a school. A filled circle already studies regularly, and the bottom two rows are the 60 in the library at 8 a.m.

The question

Does a study-skills workshop raise test scores for students at this school?

Step 1 / 12