Glossary
Polling vocabulary is used loosely almost everywhere. Here is exactly what each term means on this site.
- Axis
- One of four independent dimensions DataRadio scores: Economic, Social, Sovereignty, Force. A score runs from -100 to +100, where zero is the world's actual median rather than a designer's midpoint.
- Bloc
- A cluster found in raw answer vectors with no axis assumptions. Where blocs cut across the designed axes, the blocs are the finding.
- Effective sample size
- The sample size a weighted sample behaves like. Weighting always reduces it; DataRadio publishes it rather than the raw count alone, because the raw count overstates precision.
- Forced-choice pair
- An item made of two written statements rather than one proposition to agree with. Removes acquiescence bias, where respondents agree with whatever is put to them.
- Margin of error
- The half-width of the interval around a published figure. Printed beside every DataRadio figure; a figure quoted without it is being quoted incompletely.
- Post-stratification
- Reweighting completed responses so a sample's composition matches the population's known composition. DataRadio rakes on age, gender and education within country, then weights countries by adult population for world figures.
- Reporting threshold
- The minimum responses before a nation is broken out: 100 on the question. Below it, DataRadio publishes no national figure at all.
- Side A / Side B
- The two poles of an item. Stored values always mean 'toward side B', with the sign reversed for respondents shown the sides in the opposite order.
- Quarantine
- Responses failing integrity checks are flagged and excluded from published aggregates, never deleted. The record of what was received stays intact.
- Machine translation status
- A locale catalogue marked 'machine' has had no speaker review. DataRadio labels it on the page rather than letting a translated interface imply a review it has not had.