False Positives and False Negatives

False positives flag an absent condition, and false negatives miss a present condition. Conditional error rates, the fraction of positive results that are wrong, and overall accuracy are different quantities.

A false positive labels an absent condition as present, while a false negative misses a present condition. Their rates use different denominators: false positives among actual negatives and false negatives among actual positives. The probability that a positive result is wrong is another quantity, affected by prevalence as well as those rates.

Changing a threshold within a fixed scoring system commonly trades one error against the other, but better information or a better model can reduce both. Choose a policy using the consequences of errors, correct decisions, review capacity, and uncertainty. No error type is universally more serious across all settings or individuals.

When to use it

When analyzing data; when building predictive models; when evaluating statistical claims; when quantifying uncertainty.

How it can help

Define the target and reference outcome, calculate clearly labeled rates, and compare consequences and review capacity when choosing a threshold or improving the system.