Reading data critically
Compare groups, spot the Simpson effect, judge a sample, tell percent from percentage points and correlation from cause.
- France: grade 10
- Germany: grade 11
- Syria: grade 7
- Palestine: grade 10
- Ukraine: grade 11
Guide
Before you trust a statement, ask: who was asked or measured (the sample), how many, and was the choice random? A poll among people who chose to answer online often shows only one kind of people.
A rate such as the share of cured patients can behave differently for the whole group and for its parts. If one treatment is given mostly to severe cases, it can be better in every part and still look worse in total (the Simpson effect). Always look at the subgroups.
A change from 40 % to 80 % is +40 percentage points, but +100 % in relative terms. Say which one you mean.
Two things that go together (a correlation) need not cause each other. Often a third factor, such as the weather, explains both.