Chapter map

1. 1. Objects
2. 2. Original AB
3. 3. Original AC
4. 4. Original BC
5. 5. First merge
6. 6. Second merge
7. 7. Original vector
8. 8. Cophenetic vector
9. 9. Original deviations
10. 10. Tree deviations
11. 11. Cross-product
12. 12. Scale term
13. 13. Coefficient
14. 14. Residuals
15. 15. Cut height
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. Objects

This is where modelling choices begin to affect evidence. Use A,B,C.

A concrete situation

There are three unordered pairs.

What the mathematics does not license

Pair order must match across vectors. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Freeze order. 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. Original AB

The central mathematical idea is concrete. Set d(A,B)=1.

A concrete situation

A and B are closest.

What the mathematics does not license

The metric is part of the evidence. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Name metric. 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. Original AC

Here the geometry becomes operational. Set d(A,C)=2.

A concrete situation

This is the middle distance.

What the mathematics does not license

Keep original precision. The same reasoning helps students move from an exercise to a real dataset.

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.

---

4. Original BC

This step deserves a careful pause. Set d(B,C)=3.

A concrete situation

This is the largest distance.

What the mathematics does not license

Outliers can dominate ranges. 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.

---

5. First merge

The useful question is not only how to calculate, but what the calculation assumes. Merge A and B at height 1.

A concrete situation

Cophenetic AB=1.

What the mathematics does not license

Linkage determines merge heights. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Name linkage. 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 merge

A small example makes the mechanism visible. Join C at height 2.

A concrete situation

Cophenetic AC=BC=2.

What the mathematics does not license

A dendrogram creates an ultrametric. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show tree. 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. Original vector

This is where modelling choices begin to affect evidence. Write [1,2,3].

A concrete situation

Mean=2.

What the mathematics does not license

Use the upper triangle once. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Avoid duplicates. 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. Cophenetic vector

The central mathematical idea is concrete. Write [1,2,2].

A concrete situation

Mean=5/3.

What the mathematics does not license

Do not use leaf order as distance. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Keep vector. 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. Original deviations

Here the geometry becomes operational. Subtract two.

A concrete situation

Values [-1,0,1].

What the mathematics does not license

Centring is required for correlation. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Show terms. 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. Tree deviations

This step deserves a careful pause. Subtract 5/3.

A concrete situation

Values [-2/3,1/3,1/3].

What the mathematics does not license

Keep fractions. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Show terms. 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. Cross-product

The useful question is not only how to calculate, but what the calculation assumes. Multiply and sum.

A concrete situation

Numerator=1.

What the mathematics does not license

Pair alignment matters. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Audit order. 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. Scale term

A small example makes the mechanism visible. Compute sqrt(2 x 2/3).

A concrete situation

Denominator=1.1547.

What the mathematics does not license

Zero variance makes correlation undefined. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Check edge. 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. Coefficient

This is where modelling choices begin to affect evidence. Divide 1 by 1.1547.

A concrete situation

r=.8660.

What the mathematics does not license

Correlation measures linear fit. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Report 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.

---

14. Residuals

The central mathematical idea is concrete. Subtract cophenetic from original distances.

A concrete situation

Residuals are 0,0,1.

What the mathematics does not license

One coefficient hides which pair distorts. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Map pairs. 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. Cut height

Here the geometry becomes operational. Cut the tree at a chosen height.

A concrete situation

The coefficient does not select the cut automatically.

What the mathematics does not license

Use stability and purpose. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Separate decision. 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

This step deserves a careful pause. Name metric, linkage, pairs and ties.

A concrete situation

Teaching r=.8660.

What the mathematics does not license

Do not call it cluster truth. That distinction matters because a correct formula can still answer the wrong operational question.

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

A cophenetic correlation coefficient measures how faithfully a hierarchical clustering dendrogram preserves the original pairwise dissimilarities. For every pair of objects, the cophenetic distance is the dendrogram height at which the two objects first join the same cluster. Pearson correlation then compares the original-distance vector with the cophenetic-distance vector.

Take three objects A, B and C with original pairwise distances AB=1, AC=2 and BC=3. Suppose a dendrogram merges A and B at height 1 and then joins C at height 2. Its cophenetic distances are therefore 1, 2 and 2. The original vector is [1,2,3] and the cophenetic vector is [1,2,2].

After centring, the original deviations are [-1,0,1] and the cophenetic deviations are [-2/3,1/3,1/3]. Their dot product is 1; the square root of the product of squared-deviation sums is sqrt(2 x 2/3)=1.1547. The cophenetic correlation is 1/1.1547=.8660.

A high coefficient indicates strong linear preservation of pairwise dissimilarity order and scale in the tree, but it does not make a particular cut height correct. Ties, nonlinear distortion, outliers, the original distance metric and linkage rule all affect the result. Compare dendrograms built from the same distance data, inspect residual pair distortions and use stability and domain evidence before naming groups.

---

Frequently asked questions

What is cophenetic distance?

For a pair of objects, the dendrogram height at which they first belong to the same cluster.

What is cophenetic correlation?

Pearson correlation between original pairwise dissimilarities and their cophenetic distances.

Is higher better?

Higher linear correspondence indicates better global preservation for dendrograms built from the same original distances.

Does it choose a cut height?

No. It assesses tree fidelity and must be paired with cut-specific stability and purpose.

What can distort it?

Distance scaling, linkage, ties, outliers, nonlinear pair distortions and a few influential distances.

What must be reported?

Objects, original metric, preprocessing, linkage, pair vectors, coefficient, residual pairs, ties and cut-validation evidence.

---

Primary source and fact-check note

Robert R. Sokal and F. James Rohlf, “The Comparison of Dendrograms by Objective Methods,” Taxon 11, 33–40 (1962), DOI 10.2307/1217208. 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

• How hierarchical clustering builds dendrograms
• Dunn index and distance definitions
• Silhouette coefficient and sample geometry
• Feature scaling and distance geometry
• Mathematics World at Bukit Timah Tutor

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: clustering evidence and responsible use

Data audit

Freeze objects, features, units, exclusions, missing values, duplicates, transformations and random seeds.

Geometry audit

Document metric, scaling, linkage, sampling boundary, null reference and every pairwise or nearest-neighbour construction.

Formula audit

Show dispersion logs, Monte Carlo errors, powered distances, cophenetic pairs, co-membership counts and denominators.

Convention audit

Name gap rule, Hopkins orientation and exponent, correlation definition, classifier, threshold and weakest-cluster rule.

Uncertainty audit

Repeat reference simulations, random points, splits, seeds, bootstrap samples and plausible cluster counts.

Failure audit

Test outliers, duplicates, tiny clusters, high dimensions, correlated features, ties, leakage and boundary choices.

Meaning audit

Compare numerical tendency or stability with domain evidence without assuming groups are natural or fair.

Final claim

State exact tendency, fidelity, compactness or reproducibility evidence and stop before causal or guaranteed-usefulness claims.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Objects

Begin with this clustering mechanism: Use A,B,C. Recreate the concrete case: There are three unordered pairs. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Pair order must match across vectors. Complete the transfer step: Freeze order. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Original AB

Begin with this clustering mechanism: Set d(A,B)=1. Recreate the concrete case: A and B are closest. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: The metric is part of the evidence. Complete the transfer step: Name metric. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Original AC

Begin with this clustering mechanism: Set d(A,C)=2. Recreate the concrete case: This is the middle distance. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Keep original precision. Complete the transfer step: Record value. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Original BC

Begin with this clustering mechanism: Set d(B,C)=3. Recreate the concrete case: This is the largest distance. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Outliers can dominate ranges. Complete the transfer step: Record value. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 5: First merge

Begin with this clustering mechanism: Merge A and B at height 1. Recreate the concrete case: Cophenetic AB=1. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Linkage determines merge heights. Complete the transfer step: Name linkage. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Second merge

Begin with this clustering mechanism: Join C at height 2. Recreate the concrete case: Cophenetic AC=BC=2. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A dendrogram creates an ultrametric. Complete the transfer step: Show tree. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Original vector

Begin with this clustering mechanism: Write [1,2,3]. Recreate the concrete case: Mean=2. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Use the upper triangle once. Complete the transfer step: Avoid duplicates. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Cophenetic vector

Begin with this clustering mechanism: Write [1,2,2]. Recreate the concrete case: Mean=5/3. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not use leaf order as distance. Complete the transfer step: Keep vector. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 9: Original deviations

Begin with this clustering mechanism: Subtract two. Recreate the concrete case: Values [-1,0,1]. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Centring is required for correlation. Complete the transfer step: Show terms. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Tree deviations

Begin with this clustering mechanism: Subtract 5/3. Recreate the concrete case: Values [-2/3,1/3,1/3]. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Keep fractions. Complete the transfer step: Show terms. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Cross-product

Begin with this clustering mechanism: Multiply and sum. Recreate the concrete case: Numerator=1. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Pair alignment matters. Complete the transfer step: Audit order. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Scale term

Begin with this clustering mechanism: Compute sqrt(2 x 2/3). Recreate the concrete case: Denominator=1.1547. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Zero variance makes correlation undefined. Complete the transfer step: Check edge. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 13: Coefficient

Begin with this clustering mechanism: Divide 1 by 1.1547. Recreate the concrete case: r=.8660. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Correlation measures linear fit. Complete the transfer step: Report value. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Residuals

Begin with this clustering mechanism: Subtract cophenetic from original distances. Recreate the concrete case: Residuals are 0,0,1. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One coefficient hides which pair distorts. Complete the transfer step: Map pairs. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Cut height

Begin with this clustering mechanism: Cut the tree at a chosen height. Recreate the concrete case: The coefficient does not select the cut automatically. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Use stability and purpose. Complete the transfer step: Separate decision. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. 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 clustering mechanism: Name metric, linkage, pairs and ties. Recreate the concrete case: Teaching r=.8660. Write object IDs, original pair distances, merge heights, cophenetic distances, pair order, centred vectors, cross-products, residuals and correlation with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call it cluster truth. Complete the transfer step: Bound conclusion. Change exactly one feature scale, cluster count, null boundary, simulation replicate, exponent, random location, nearest neighbour, distance, linkage, merge height, test split, classifier label, cluster size, threshold or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep tendency, compactness, tree fidelity, reproducibility, uncertainty and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving a uniform cloud, a correlated box, one outlier, tied distances, a tiny cluster or leaked preprocessing. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.