Equivalence test
A test that asks not 'is there an effect?' but 'how big could an effect be and still hide in this data?', so that a null result can actually rule out large effects.
In plain English
An ordinary test that finds nothing only says it didn't find anything: the effect might exist but be too small for the data to see. An equivalence test sets a limit in advance, say 2 percentage points, and checks whether the data rule out anything bigger. If they do, you can say more than 'no evidence'. You can say 'no effect larger than this'.
Uncle Gary explains
Uncle Gary says the team kicking first wins 60% of penalty shootouts. A study of about 7,000 shootouts ruled out any first-kicker edge bigger than 1.8 percentage points. Gary says that's the most depressing thing ever said in the Red Lion.
Why it matters for your money
'We found no edge' and 'there is no edge' are different claims. An equivalence test is how you get from the first towards the second, and how you tell a strategy that failed a test from one the test was too small to judge.
Where we tested it
Related terms
All glossary terms. Educational content, not financial advice.