3 ms·
>Alternative index value thresholds for positivity resulted in 100% sensitivity and 100% specificity in this cohort. This would be true of a coin flip, wouldn'
by whatshisface 7y ago
>Alternative index value thresholds for positivity resulted in 100% sensitivity and 100% specificity in this cohort.
This would be true of a coin flip, wouldn't it? "Head means positive, tails means not negative" would be 100% sensitivity and "Tails means negative, heads means you're not positive" would be 100% specificity. So those would be the "alternative index value thresholds."
https://en.wikipedia.org/wiki/Sensitivity_and_specificity https://en.wikipedia.org/wiki/Sensitivity_and_specificity
- trehalose 7y agoYou misunderstand. A test that labels everyone positive would be 100% sensitivity but 0% specificity, and a test that labels everyone negative would be 100% specificity but 0% sensitivity. Deciding every case at once by a single coin flip would thus be useless. As the article says, "A perfect predictor would be described as 100% sensitive, meaning all sick individuals are correctly identified as sick, and 100% specific, meaning no healthy individuals are incorrectly identified as sick." Deciding every case at once with a single coin flip does not meet both critera of a perfect predictor.
- whatshisface 7y agoNot at the same time, but it does if you split it in to two separate tests, each with an "Alternative index value threshold."