Decision Threshold
Definition
The cut-off score above which a classification method declares a positive result. Raising the threshold reduces FPR but increases FNR; lowering it does the reverse. The chosen threshold must be documented and justified for each operational context.
- Function
- Cut-off score for declaring a positive classification result
- Raise effect
- Reduces false positive rate, increases false negative rate
- Lower effect
- Reduces false negative rate, increases false positive rate
- Requirement
- Must be documented and justified for the operational context
- Field
- Error-rate evaluation across forensic classification methods
Common questions
Why can't a decision threshold be set to minimise both false positives and false negatives at once?+
The two error rates trade off against each other along the same underlying score distribution, so tightening the threshold to catch fewer false positives inevitably lets more true positives slip past as false negatives, and vice versa. There is no threshold that minimises both simultaneously.
How should the appropriate threshold differ between a screening test and a confirmatory test?+
A screening test is usually set with a lower threshold to minimise false negatives, since missed positives carry a higher cost at that stage, while a confirmatory test that follows it can use a stricter threshold because its role is to rule out the false positives the screen let through.
Related terms
- False Negative Rate (FNR)
- The proportion of true matches (or true positives) that a classification method fails to detect, declaring them as non-matches. Also called the...
- False Positive Rate (FPR)
- The proportion of true non-matches (or true negatives) that a classification method declares as matches (or positives). Also called the type I...
- ROC Curve
- Receiver operating characteristic curve: a plot of sensitivity (y-axis) against FPR (x-axis) as the decision threshold is swept from its most lenient...
- Sensitivity (True Positive Rate)
- The proportion of true matches correctly identified as matches. Sensitivity = TP / (TP + FN) = 1 minus FNR. A method...
- Specificity (True Negative Rate)
- The proportion of true non-matches correctly identified as non-matches. Specificity = TN / (TN + FP) = 1 minus FPR. A method...