WHY MATHEMATICS?

Why Mathematics? | ROC AUC for Binary Predictions: Pairwise Ranking, Ties, Thresholds and Class Prevalence

How AUC counts positive-negative score pairs, with a complete four-pair example, threshold geometry, limitations and practical guidance.

Why mathematics is important becomes easier to see when a real decision depends on it. This guide follows the mechanism from definitions and dimensions to a worked calculation, validation, practical learning and responsible use. It keeps factual evidence separate from interpretation and treats every illustrative scenario as an example rather than a promise of results.

Choose a reading route

Build ROC operating points from binary score thresholds
Interpret AUC as a pairwise ranking probability
Handle ties and uncertainty transparently
Separate discrimination from calibration and utility

Chapter map

1. Positive class
2. Scores
3. Negative scores
4. Pairs
5. First positive
6. Second positive
7. AUC
8. Tie
9. ROC thresholds
10. TPR
11. FPR
12. Transformation
13. Prevalence
14. Partial range
15. Uncertainty
16. Claim

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

1. Positive class

A small example makes the mechanism visible. Define the event of interest.

A concrete situation

Code positives consistently.

What the mathematics does not license

Reversing the class reverses score direction. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Freeze target. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

2. Scores

This is where modelling choices begin to affect evidence. Use positive scores .90,.60.

A concrete situation

Higher means more positive.

What the mathematics does not license

Probabilities are not required for ranking. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

State scale. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

3. Negative scores

The central mathematical idea is concrete. Use .70,.20.

A concrete situation

Keep the same score direction.

What the mathematics does not license

Case mix affects uncertainty. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record sample. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

4. Pairs

Here the geometry becomes operational. Form 2×2=4 positive-negative pairs.

A concrete situation

Count every cross-class pair.

What the mathematics does not license

Within-class order does not enter AUC. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Show denominator. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

Did You Know? In this chapter, the most important number is useful only together with its construction rule. Changing scaling, sample membership or validation design can change the numerical answer without changing the arithmetic.

5. First positive

This step deserves a careful pause. .90 beats .70 and .20.

A concrete situation

Award two wins.

What the mathematics does not license

Equality needs a tie rule. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Audit comparisons. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

6. Second positive

The useful question is not only how to calculate, but what the calculation assumes. .60 beats .20 but not .70.

A concrete situation

Award one win.

What the mathematics does not license

Do not choose thresholds yet. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Keep ranking. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

7. AUC

A small example makes the mechanism visible. Divide three wins by four pairs.

A concrete situation

AUC=.75.

What the mathematics does not license

This is empirical and sample-specific. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Report interval. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

8. Tie

This is where modelling choices begin to affect evidence. Award .5 for equal scores if using the standard convention.

A concrete situation

State tie handling.

What the mathematics does not license

Hidden rules change AUC. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Document. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

9. ROC thresholds

The central mathematical idea is concrete. Move threshold through unique scores.

A concrete situation

Calculate TPR and FPR.

What the mathematics does not license

Interpolation conventions matter visually. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

List points. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

10. TPR

Here the geometry becomes operational. Use TP/(TP+FN).

A concrete situation

Name sensitivity.

What the mathematics does not license

Undefined if no positives. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Check counts. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

11. FPR

This step deserves a careful pause. Use FP/(FP+TN).

A concrete situation

Use negative denominator.

What the mathematics does not license

Do not confuse with one-specificity without coding check. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Audit. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

12. Transformation

The useful question is not only how to calculate, but what the calculation assumes. Apply a strictly increasing score transform.

A concrete situation

Ranking and AUC stay fixed.

What the mathematics does not license

Calibration can change. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Explain invariance. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

13. Prevalence

A small example makes the mechanism visible. Change positive proportion while holding class score distributions.

A concrete situation

AUC target may remain similar.

What the mathematics does not license

Decision values and predictive values change. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Add context. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

14. Partial range

This is where modelling choices begin to affect evidence. Inspect operational FPR region.

A concrete situation

Report partial or threshold metrics if justified.

What the mathematics does not license

Global AUC can hide local weakness. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Predeclare region. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

15. Uncertainty

The central mathematical idea is concrete. Use paired methods for model comparisons.

A concrete situation

Preserve correlated predictions.

What the mathematics does not license

Independent intervals can misstate differences. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Report design. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

16. Claim

Here the geometry becomes operational. Name event, score direction and ties.

A concrete situation

Teaching AUC is .75.

What the mathematics does not license

Do not call AUC calibration or utility. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Bound conclusion. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

A seven-day practice lab

Day 1: define the decision, observation unit, X variables and Y or reference condition.
Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
Day 3: build a transparent baseline and write what it assumes.
Day 4: fit the main method inside a training-only pipeline; record every parameter.
Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
Day 6: compare validation metrics with the practical workload created by errors.
Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

Which part of the model is a mathematical definition, and which part is a choice?
What small calculation would catch a sign, scale or dimension error?
What evidence comes from training, validation and the untouched test?
What benign condition could look like a fault or prediction failure?
How would the answer change for another response, asset, school project or time period?
What should a human review before acting on the score?

A method-specific audit before deployment

A receiver operating characteristic curve follows a binary ranking score across every possible decision threshold. At each threshold it plots the true-positive rate against the false-positive rate. The area under the curve, AUC, has a useful pairwise interpretation: it is the probability that a randomly chosen positive case receives a higher score than a randomly chosen negative case, with ties commonly receiving half credit.

Take positive-case scores .90 and .60 and negative-case scores .70 and .20. There are four positive-negative pairs. The .90 score beats both negatives, while .60 beats .20 but loses to .70. Three of four pairs are correctly ordered, so empirical AUC=3/4=.75. A tie would contribute .5 under the stated convention.

AUC measures discrimination, not probability calibration. Any strictly increasing transformation of scores preserves ranks and therefore preserves AUC while changing predicted probabilities. AUC can also hide performance in the threshold range that matters, and its apparent stability across class prevalence does not mean decisions or predictive values are prevalence-free.

A responsible analysis reports positive-class definition, score direction, ties, sample composition, paired uncertainty and threshold-specific sensitivity and specificity. It examines calibration and decision consequences separately. Students can connect inequalities, pair counting, geometry and integration to the mathematics of ranking systems used in science, engineering and AI.

Frequently asked questions

What is ROC AUC?

Area under the true-positive-rate versus false-positive-rate curve across score thresholds.

What is its pairwise meaning?

The probability a random positive is scored above a random negative, with a stated tie convention.

Does AUC measure calibration?

No. Monotone score transformations preserve AUC while changing probabilities.

Does prevalence never matter?

AUC’s pairwise target can be prevalence-invariant, while uncertainty, predictive values and decisions still depend on case mix.

Why inspect thresholds?

A global area can hide weakness in the operating region that matters.

What must be reported?

Class, score direction, ties, counts, curve construction, uncertainty, thresholds, calibration and utility.

Primary source and fact-check note

James A. Hanley and Barbara J. McNeil, “The Meaning and Use of the Area under a Receiver Operating Characteristic (ROC) Curve,” Radiology 143(1), 29–36 (1982), DOI 10.1148/radiology.143.1.7063747. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Extended laboratory: decision-focused validation and responsible use

Target audit

Freeze event, positive class, score direction, case set, action and consequence model before evaluation.

Formula audit

Write every pair count, threshold count, odds weight, cost, loss, expense, reference and denominator explicitly.

Sample audit

Report positives, negatives, prevalence, event rate, missing cases, weights and dependence.

Threshold audit

Choose operating regions from real decisions before inspecting results; do not optimise on the evaluation outcomes.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Reference audit

Compare model curves with transparent no-skill, treat-all, treat-none, climatology or perfect-information references as appropriate.

Calibration audit

Add probability calibration evidence because ranking and threshold metrics can look favourable with wrong probabilities.

Uncertainty audit

Use paired, dependence-aware uncertainty and sensitivity analysis for scores, curves and consequences.

Final claim

State the exact ranking or conditional-value evidence and stop before universal, causal, salary or guaranteed-outcome claims.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Positive class

Begin with this mechanism: Define the event of interest. Recreate the concrete case: Code positives consistently. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Reversing the class reverses score direction. Complete the transfer step: Freeze target. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 2: Scores

Begin with this mechanism: Use positive scores .90,.60. Recreate the concrete case: Higher means more positive. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Probabilities are not required for ranking. Complete the transfer step: State scale. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 3: Negative scores

Begin with this mechanism: Use .70,.20. Recreate the concrete case: Keep the same score direction. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Case mix affects uncertainty. Complete the transfer step: Record sample. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 4: Pairs

Begin with this mechanism: Form 2×2=4 positive-negative pairs. Recreate the concrete case: Count every cross-class pair. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Within-class order does not enter AUC. Complete the transfer step: Show denominator. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 5: First positive

Begin with this mechanism: .90 beats .70 and .20. Recreate the concrete case: Award two wins. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Equality needs a tie rule. Complete the transfer step: Audit comparisons. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 6: Second positive

Begin with this mechanism: .60 beats .20 but not .70. Recreate the concrete case: Award one win. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not choose thresholds yet. Complete the transfer step: Keep ranking. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 7: AUC

Begin with this mechanism: Divide three wins by four pairs. Recreate the concrete case: AUC=.75. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: This is empirical and sample-specific. Complete the transfer step: Report interval. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 8: Tie

Begin with this mechanism: Award .5 for equal scores if using the standard convention. Recreate the concrete case: State tie handling. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Hidden rules change AUC. Complete the transfer step: Document. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 9: ROC thresholds

Begin with this mechanism: Move threshold through unique scores. Recreate the concrete case: Calculate TPR and FPR. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Interpolation conventions matter visually. Complete the transfer step: List points. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 10: TPR

Begin with this mechanism: Use TP/(TP+FN). Recreate the concrete case: Name sensitivity. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Undefined if no positives. Complete the transfer step: Check counts. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 11: FPR

Begin with this mechanism: Use FP/(FP+TN). Recreate the concrete case: Use negative denominator. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not confuse with one-specificity without coding check. Complete the transfer step: Audit. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 12: Transformation

Begin with this mechanism: Apply a strictly increasing score transform. Recreate the concrete case: Ranking and AUC stay fixed. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Calibration can change. Complete the transfer step: Explain invariance. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 13: Prevalence

Begin with this mechanism: Change positive proportion while holding class score distributions. Recreate the concrete case: AUC target may remain similar. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Decision values and predictive values change. Complete the transfer step: Add context. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 14: Partial range

Begin with this mechanism: Inspect operational FPR region. Recreate the concrete case: Report partial or threshold metrics if justified. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Global AUC can hide local weakness. Complete the transfer step: Predeclare region. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 15: Uncertainty

Begin with this mechanism: Use paired methods for model comparisons. Recreate the concrete case: Preserve correlated predictions. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Independent intervals can misstate differences. Complete the transfer step: Report design. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.

Workshop 16: Claim

Begin with this mechanism: Name event, score direction and ties. Recreate the concrete case: Teaching AUC is .75. Write class labels, scores, cross-class pairs, ties, thresholds, TPR, FPR, AUC and uncertainty with signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call AUC calibration or utility. Complete the transfer step: Bound conclusion. Change exactly one class label, score, tie, threshold, prevalence, ranked case, positive count, AP convention, true positive, false positive, threshold probability, cost, loss, base rate, action effectiveness or validation split. Explain which mathematical object changes, which curve or value moves and which claim must weaken. Keep target, arithmetic, uncertainty and interpretation separate. Explain once to a student and once to a parent. Add a counterexample involving unchanged AUC with bad calibration, high AP under shifted prevalence, optimistic in-sample net benefit or an attractive cost-loss value under unrealistic full protection. State the evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one next measurement.