Saturation
In measurement theory, the loss of a test's ability to discriminate among test takers. When all model scores approach a ceiling or floor effect, variance becomes minimal and the test can no longer provide valid information about differences in ability.
Timeline (last 90 days)
Saturation in measurement theory refers to test scores losing the ability to discriminate between the subjects being measured; when all models' scores approach the ceiling or floor, the variance is extremely small.
If a saturated dataset is still included in the aggregate total score, its effect is equivalent to mixing a constant into the average score, which not only contributes no information but also dilutes the signal weight of other datasets that still have discriminating power.
All Facts (2)
Saturation in measurement theory refers to test scores losing the ability to discriminate between the subjects being measured; when all models' scores approach the ceiling or floor, the variance is extremely small.
50%UnverifiedIf a saturated dataset is still included in the aggregate total score, its effect is equivalent to mixing a constant into the average score, which not only contributes no information but also dilutes the signal weight of other datasets that still have discriminating power.
50%