The reason a statistic needs a comparison to mean anything
A number alone is a fact without a frame. The comparison supplied beside it is what gives it size, direction and significance.
Filed by The Archivist 2 min read
Intuition test — answer before you read on
A report says the error rate is three per cent. When does this number mean something?
Correct answer: B
Kahneman and Miller show that evaluation requires a norm, and the norm is constructed from the available comparison. Methodology and sample size matter, but a number without a comparison has no direction.
A report says the error rate is three per cent. Without a comparison, the number floats: is three per cent good or bad? Compared to last year’s five per cent it is progress. Compared to a competitor’s one per cent it is a problem. The number is the same; the comparison determines the conclusion.
What everyone sees
A statistic looks like a complete piece of information: a number plus a unit should be enough. The assumption is that the reader can evaluate a number on its own, the way they can evaluate a photograph without a caption.
What is actually happening
Kahneman and Miller’s norm theory holds that every stimulus is evaluated against a norm, and the norm is constructed from whatever comparison is available. Best described the use and misuse of statistics in public discourse and showed that the same number can support opposite conclusions depending on which comparison is offered. Tufte argued that every data display is a comparison and that a number presented without one leaves the reader to supply their own, which may be the wrong one.
Why it stays hidden
The incompleteness hides because the number carries a sense of precision that feels like completeness. Three per cent sounds like a fact, and facts are supposed to be self-sufficient. The comparison that gives the fact its meaning is chosen by the presenter and is not marked as a choice — it is presented as context, which sounds neutral, and context controls the conclusion.
A number alone is not an argument. The comparison beside it is the argument, and the person who chose the comparison chose the conclusion.
A number alone is not an argument. The comparison beside it is the argument, and the person who chose the comparison chose the conclusion.
Collect this card
A number alone is not an argument. The comparison beside it is the argument, and the person who chose the comparison chose the conclusion.
0 / 10,000 collected
Sources & further reading 3
- Kahneman & Miller, "Norm Theory: Comparing Reality to Its Alternatives", Psychological Review, 1986
- Best, "Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists", 2001
- Tufte, "The Visual Display of Quantitative Information", 2001
Cross-references
Related files
Filed near this one in the index.
-
No visual on fileNarrative Structure Entry #0530
Why an unresolved ending is remembered longer
An interrupted task stays in mind longer than a completed one, which is why unresolved endings outlive the stories that finish tidily.
MasterThe hidden part #0530Completion releases attention. Anything left open keeps a claim on it.
Narrative Open file -
Hidden Words Entry #0001
Why “clue” used to be a ball of yarn
Before it meant evidence, a clue was a physical thread — the one Theseus unwound through the labyrinth so he could find the way back out.
MasterThe hidden part #0001Following a clue is not a metaphor for detection. It is a description of walking backwards along string.
Words Open file -
No visual on fileNarrative Structure Entry #0535
The reason testimony beats aggregate evidence
One person’s account defeats a table of ten thousand. The account has a face, a sequence and a cost. The table has none of these.
MasterThe hidden part #0535A story activates a path the table cannot reach. The table has magnitude; the story has a face. The face wins, and this is a finding about the audience.
Narrative Open file -
No visual on fileNaming & Sound Entry #0560
The reason a new category needs a new word
Without a name the new thing is described by analogy to old things. The analogy supplies expectations the new thing cannot meet, and it fails before it is understood.
MasterThe hidden part #0560A name is not a label. It is a category boundary. Without it the new thing is judged by standards it was never meant to meet.
Naming Open file
Annotations are reserved for archive members.
Sign in to annotate