Chapter map

1. 1. Candidate range
2. 2. Alpha
3. 3. Scatter definition
4. 4. Separation definition
5. 5. Two scatter
6. 6. Two separation
7. 7. SD two
8. 8. Three scatter
9. 9. Three separation
10. 10. SD three
11. 11. Four scatter
12. 12. Four separation
13. 13. SD four
14. 14. Selection
15. 15. Coincident centres
16. 16. Claim

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

---

1. Candidate range

The useful question is not only how to calculate, but what the calculation assumes. Use k=2,3,4 with cmax=4.

A concrete situation

The maximum candidate defines alpha.

What the mathematics does not license

Changing range changes the criterion. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

State range. 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. Alpha

A small example makes the mechanism visible. Set Dis(4)=.55.

A concrete situation

Therefore alpha=.55.

What the mathematics does not license

Do not refit alpha separately for each k. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record alpha. 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. Scatter definition

This is where modelling choices begin to affect evidence. Use average cluster scatter relative to global scatter.

A concrete situation

Smaller means more compact clusters.

What the mathematics does not license

Variance norm must be named. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

State formula. 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. Separation definition

The central mathematical idea is concrete. Use centre-distance ratio and reciprocal distance sums.

A concrete situation

Smaller is preferred in SD.

What the mathematics does not license

Near-coincident centres can dominate. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

State formula. 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.

---

5. Two scatter

Here the geometry becomes operational. Set Scat2=.25.

A concrete situation

Two clusters retain moderate spread.

What the mathematics does not license

Values are dimensionless ratios. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record value. 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. Two separation

This step deserves a careful pause. Set Dis2=.70.

A concrete situation

This is the declared centre geometry term.

What the mathematics does not license

Check Dmin is nonzero. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record value. 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. SD two

The useful question is not only how to calculate, but what the calculation assumes. Compute .55x.25+.70.

A concrete situation

SD2=.8375.

What the mathematics does not license

Lower is better. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show step. 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. Three scatter

A small example makes the mechanism visible. Set Scat3=.18.

A concrete situation

Compactness improves.

What the mathematics does not license

Use identical preprocessing. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record value. 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. Three separation

This is where modelling choices begin to affect evidence. Set Dis3=.50.

A concrete situation

Centre geometry improves.

What the mathematics does not license

Same distance metric is required. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record value. 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. SD three

The central mathematical idea is concrete. Compute .55x.18+.50.

A concrete situation

SD3=.599.

What the mathematics does not license

Keep alpha fixed. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show step. 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. Four scatter

Here the geometry becomes operational. Set Scat4=.16.

A concrete situation

Compactness improves slightly.

What the mathematics does not license

More clusters often reduce scatter. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Record value. 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. Four separation

This step deserves a careful pause. Use Dis4=.55.

A concrete situation

This is also alpha.

What the mathematics does not license

Candidate four worsens the geometry term. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record value. 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. SD four

The useful question is not only how to calculate, but what the calculation assumes. Compute .55x.16+.55.

A concrete situation

SD4=.638.

What the mathematics does not license

The minimum is not terminal. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show step. 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. Selection

A small example makes the mechanism visible. Compare .8375,.599,.638.

A concrete situation

Three clusters win.

What the mathematics does not license

Repeat under plausible ranges. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

State minimum. 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. Coincident centres

This is where modelling choices begin to affect evidence. Let Dmin approach zero.

A concrete situation

Dis can explode.

What the mathematics does not license

Merge, diagnose or reject the fit. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Flag instability. 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

The central mathematical idea is concrete. Name alpha,cmax,scatter norm,centre distances and range.

A concrete situation

Teaching minimum is k=3.

What the mathematics does not license

Do not imply useful categories. The same reasoning helps students move from an exercise to a real dataset.

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

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

The SD validity index combines average within-cluster scatter with a centre-separation term. A common form is SD(k)=alpha Scat(k)+Dis(k), where alpha is fixed as Dis(cmax), the separation value at the largest candidate cluster count. Lower values are preferred.

Let the candidate range end at cmax=4 and suppose Dis(4)=0.55, so alpha=0.55. For two clusters, let Scat(2)=0.25 and Dis(2)=0.70. Then SD(2)=0.55×0.25+0.70=0.8375.

For three clusters, let Scat(3)=0.18 and Dis(3)=0.50, giving SD(3)=0.55×0.18+0.50=0.599. For four clusters, let Scat(4)=0.16 and Dis(4)=0.55, giving SD(4)=0.55×0.16+0.55=0.638. The three-cluster candidate is the minimum.

The Dis term uses the maximum-to-minimum centre-distance ratio and reciprocal sums of distances from each centre to the others. A near-zero minimum centre distance can make it explode. Because alpha comes from cmax, changing the candidate range can change every SD value. Scale, centre definition, variance norm and range are part of the evidence.

---

Frequently asked questions

What is the SD validity index?

An internal criterion adding a weighted average-scatter term to a centre-separation geometry term.

Which direction is better?

Lower SD values are preferred.

What is alpha?

In the stated convention, alpha is Dis(cmax), fixed from the largest candidate cluster count.

Why does the candidate range matter?

Changing cmax changes alpha and therefore changes every candidate score.

What happens when two centres nearly coincide?

Dmin approaches zero and the separation term can become unstable or explode.

What must be reported?

Scatter norm, centre distance, Dmax, Dmin, Dis values, cmax, alpha, candidate range, scaling, algorithm, starts and sensitivity checks.

---

Primary source and fact-check note

Maria Halkidi, Yannis Batistakis and Michalis Vazirgiannis, “On Clustering Validation Techniques,” Journal of Intelligent Information Systems 17(2–3), 107–145 (2001), DOI 10.1023/A:1012801612483. 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

• S_Dbw density-aware extension
• Davies–Bouldin scatter and centre separation
• Calinski–Harabasz variance ratio
• How AI finds patterns without labels
• How Mathematics Works: Statistics

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: cluster-validity evidence and responsible use

Data audit

Freeze observations, variables, units, exclusions, missing-data rules, standardisation, sample size and seeds.

Partition audit

Use comparable algorithms, distances, cluster-count ranges, stopping rules and random-start budgets.

Formula audit

Show variance ratios, pair counts, sorted distances, reference sums, distance means, balance factors, scatter norms and separation terms.

Convention audit

Name aggregation, pair coding, standard-deviation divisor, ordered or unordered pairs, alpha definition and higher-or-lower direction.

Uncertainty audit

Repeat starts, samples, feature scales, distances, candidate ranges and plausible implementation conventions.

Failure audit

Test constant variables, distance ties, zero reference ranges, zero distance spread, singletons, imbalanced groups and coincident centres.

Meaning audit

Compare internal numerical evidence with stability and domain evidence without assuming clusters are natural categories.

Final claim

State exact geometric evidence and stop before causal, semantic, career or guaranteed-performance claims.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Candidate range

Begin with this cluster-validity mechanism: Use k=2,3,4 with cmax=4. Recreate the concrete case: The maximum candidate defines alpha. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Changing range changes the criterion. Complete the transfer step: State range. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Alpha

Begin with this cluster-validity mechanism: Set Dis(4)=.55. Recreate the concrete case: Therefore alpha=.55. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not refit alpha separately for each k. Complete the transfer step: Record alpha. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Scatter definition

Begin with this cluster-validity mechanism: Use average cluster scatter relative to global scatter. Recreate the concrete case: Smaller means more compact clusters. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Variance norm must be named. Complete the transfer step: State formula. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Separation definition

Begin with this cluster-validity mechanism: Use centre-distance ratio and reciprocal distance sums. Recreate the concrete case: Smaller is preferred in SD. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Near-coincident centres can dominate. Complete the transfer step: State formula. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 5: Two scatter

Begin with this cluster-validity mechanism: Set Scat2=.25. Recreate the concrete case: Two clusters retain moderate spread. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Values are dimensionless ratios. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Two separation

Begin with this cluster-validity mechanism: Set Dis2=.70. Recreate the concrete case: This is the declared centre geometry term. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Check Dmin is nonzero. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: SD two

Begin with this cluster-validity mechanism: Compute .55x.25+.70. Recreate the concrete case: SD2=.8375. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Lower is better. Complete the transfer step: Show step. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Three scatter

Begin with this cluster-validity mechanism: Set Scat3=.18. Recreate the concrete case: Compactness improves. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Use identical preprocessing. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 9: Three separation

Begin with this cluster-validity mechanism: Set Dis3=.50. Recreate the concrete case: Centre geometry improves. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Same distance metric is required. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: SD three

Begin with this cluster-validity mechanism: Compute .55x.18+.50. Recreate the concrete case: SD3=.599. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Keep alpha fixed. Complete the transfer step: Show step. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Four scatter

Begin with this cluster-validity mechanism: Set Scat4=.16. Recreate the concrete case: Compactness improves slightly. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: More clusters often reduce scatter. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Four separation

Begin with this cluster-validity mechanism: Use Dis4=.55. Recreate the concrete case: This is also alpha. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Candidate four worsens the geometry term. Complete the transfer step: Record value. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 13: SD four

Begin with this cluster-validity mechanism: Compute .55x.16+.55. Recreate the concrete case: SD4=.638. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: The minimum is not terminal. Complete the transfer step: Show step. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Selection

Begin with this cluster-validity mechanism: Compare .8375,.599,.638. Recreate the concrete case: Three clusters win. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Repeat under plausible ranges. Complete the transfer step: State minimum. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Coincident centres

Begin with this cluster-validity mechanism: Let Dmin approach zero. Recreate the concrete case: Dis can explode. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Merge, diagnose or reject the fit. Complete the transfer step: Flag instability. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 16: Claim

Begin with this cluster-validity mechanism: Name alpha,cmax,scatter norm,centre distances and range. Recreate the concrete case: Teaching minimum is k=3. Write candidate k, cluster scatter norms, global scatter norm, centre distances, Dmax, Dmin, Dis, cmax, alpha and SD with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not imply useful categories. Complete the transfer step: Bound conclusion. Change exactly one variable, scale, explained-variance ratio, cluster count, pair count, distance, reference sum, distance spread, centre location, candidate range, alpha, initialisation or seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep compactness, separation, balance, stability, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a constant variable, tied distances, collapsed reference range, zero standard deviation, singleton, imbalanced cluster or coincident centres. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.