Chapter map

1. 1. Data
2. 2. Starting location
3. 3. Scale
4. 4. Tuning constant
5. 5. Standardize
6. 6. Inside weight
7. 7. Outside weight
8. 8. Weighted numerator
9. 9. Weighted denominator
10. 10. One-step update
11. 11. Central root
12. 12. Remote root
13. 13. Efficiency claim
14. 14. Leverage
15. 15. Uncertainty
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. Data

The central mathematical idea is concrete. Freeze observations, units and missing-value rules.

A concrete situation

Use 1,2,3,4,20.

What the mathematics does not license

A second cluster may be genuine. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Inspect 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.

---

2. Starting location

Here the geometry becomes operational. Choose a robust initial centre.

A concrete situation

Begin at m=3.

What the mathematics does not license

Starting at 20 can select another root. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record start. 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. Scale

This step deserves a careful pause. Choose a positive robust scale.

A concrete situation

Use s=1 only for the teaching calculation.

What the mathematics does not license

Scale controls cutoff membership. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Report estimator. 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. Tuning constant

The useful question is not only how to calculate, but what the calculation assumes. Choose c for a documented efficiency–resistance goal.

A concrete situation

Use c=4.685.

What the mathematics does not license

It is not a raw-data threshold. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

State model. 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. Standardize

A small example makes the mechanism visible. Compute u_i=(x_i−m)/s.

A concrete situation

Initial u values are −2,−1,0,1,17.

What the mathematics does not license

Wrong units corrupt every weight. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Check dimensions. 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. Inside weight

This is where modelling choices begin to affect evidence. Calculate [1−(u/c)²]² for |u|<c.

A concrete situation

Weights are about .66873,.91096,1,.91096.

What the mathematics does not license

Weights are nonlinear. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Keep digits. 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. Outside weight

The central mathematical idea is concrete. Set weight zero beyond c.

A concrete situation

The value 20 receives zero initially.

What the mathematics does not license

Zero weight is model-specific evidence. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Investigate point. 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. Weighted numerator

Here the geometry becomes operational. Sum w_ix_i.

A concrete situation

Keep each contribution visible.

What the mathematics does not license

Hidden deletion cannot be audited. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show table. 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. Weighted denominator

This step deserves a careful pause. Sum all positive weights.

A concrete situation

The denominator exceeds three.

What the mathematics does not license

A tiny denominator signals a failed fit. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Monitor it. 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. One-step update

The useful question is not only how to calculate, but what the calculation assumes. Divide numerator by denominator.

A concrete situation

m_1≈2.61684.

What the mathematics does not license

One step is not convergence. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Continue iterating. 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. Central root

A small example makes the mechanism visible. Iterate from the robust start.

A concrete situation

The symmetric central observations lead to 2.5.

What the mathematics does not license

Solver tolerance affects final digits. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record path. 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. Remote root

This is where modelling choices begin to affect evidence. Search alternative starting regions.

A concrete situation

m=20 is also a zero-score root with fixed s=1.

What the mathematics does not license

Non-convexity creates multiple solutions. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Use multi-start. 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. Efficiency claim

The central mathematical idea is concrete. Tie c≈4.685 to the normal-model setting.

A concrete situation

Nominal efficiency is about 95% for the usual M procedure.

What the mathematics does not license

Efficiency is model-dependent. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Cite source. 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. Leverage

Here the geometry becomes operational. In regression inspect x-space leverage.

A concrete situation

A small residual can coexist with high leverage.

What the mathematics does not license

Residual weighting alone is incomplete. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use robust design method. 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

This step deserves a careful pause. Choose a covariance or resampling method compatible with the fit.

A concrete situation

Report centre and uncertainty together.

What the mathematics does not license

Naive least-squares SE can mislead. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Validate coverage. 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 useful question is not only how to calculate, but what the calculation assumes. State the chosen root, start, scale and c.

A concrete situation

The central teaching solution is 2.5.

What the mathematics does not license

Do not call zero-weight observations errors. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Keep scope. 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

Tukey’s bisquare, also called the biweight, uses ψ(u)=u[1−(u/c)²]² for |u|<c and ψ(u)=0 beyond the cutoff. In iteratively reweighted form, w(u)=ψ(u)/u=[1−(u/c)²]² inside the cutoff and zero outside. Small standardized residuals receive weights near one, moderate residuals fade smoothly and sufficiently remote residuals contribute no score.

For 1,2,3,4,20, start at m=3 with fixed teaching scale s=1 and c=4.685. The one-step weights are approximately .66873,.91096,1,.91096,0, giving a weighted update m≈2.61684. Continuing from the central start reaches the central fixed point 2.5. The value 20 is also an isolated zero-score root when every other residual lies beyond the cutoff, revealing why initialization cannot be an afterthought.

The familiar c≈4.685 is chosen for about 95% asymptotic efficiency in a normal-model M-estimation setting, not because 4.685 is a universal outlier boundary. Residuals must first be divided by a defensible robust scale. Changing the scale changes which points cross the cutoff, and changing c changes both efficiency and protection.

A redescending criterion is non-convex and may have several local minima or estimating-equation roots. A robust high-breakdown starting fit, explicit convergence checks and inspection of alternative starts are essential. Zero weight does not prove an observation is erroneous; it says the selected fitted mechanism chose not to let that standardized residual influence this solution.

---

Frequently asked questions

What is Tukey’s bisquare score?

u[1−(u/c)²]² inside |u|<c and zero outside.

Why is it called redescending?

Its score returns to zero as a standardized residual reaches the cutoff.

What does c≈4.685 mean?

A common tuning for about 95% normal-model asymptotic efficiency in the usual M-estimation setting.

Why can there be several roots?

The bounded flat-tail loss is non-convex.

Does zero weight prove an error?

No. It records the influence decision of one fitted mechanism.

What must be reported?

Starting fit, scale, c, update rule, selected root, convergence checks and uncertainty method.

---

Primary source and fact-check note

Albert E. Beaton and John W. Tukey, “The Fitting of Power Series, Meaning Polynomials, Illustrated on Band-Spectroscopic Data,” Technometrics 16(2), 147–185 (1974), DOI 10.1080/00401706.1974.10489171. 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

• Rousseeuw–Croux Qn robust scale
• Gel–Gastwirth robust moment test
• How Robust Statistics Work
• How Mathematics Works

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: topic-specific validation and responsible use

Design audit

Freeze observations, sampling units, dependence, missingness and the target parameter.

Scale audit

State the residual scale, its units, its robustness and whether it is recomputed.

Score audit

Show every standardized residual, threshold, score, weight and contribution.

Tuning audit

Connect confidence tuning or rejection constants to a stated model and goal.

Solver audit

Record initialization, bracketing or IRLS updates, tolerance, objective and all plausible roots.

Graph audit

Plot raw data, residuals, score curves, weights and start-to-solution paths.

Sensitivity audit

Change one variance input, scale, tuning constant, starting fit or cluster at a time.

Final claim

Keep the conclusion inside the sampling design and the protection the selected score actually supplies.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Data

Begin with this mechanism: Freeze observations, units and missing-value rules. Recreate the concrete case: Use 1,2,3,4,20. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A second cluster may be genuine. Complete the transfer step: Inspect context. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Starting location

Begin with this mechanism: Choose a robust initial centre. Recreate the concrete case: Begin at m=3. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Starting at 20 can select another root. Complete the transfer step: Record start. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Scale

Begin with this mechanism: Choose a positive robust scale. Recreate the concrete case: Use s=1 only for the teaching calculation. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Scale controls cutoff membership. Complete the transfer step: Report estimator. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Tuning constant

Begin with this mechanism: Choose c for a documented efficiency–resistance goal. Recreate the concrete case: Use c=4.685. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: It is not a raw-data threshold. Complete the transfer step: State model. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 5: Standardize

Begin with this mechanism: Compute u_i=(x_i−m)/s. Recreate the concrete case: Initial u values are −2,−1,0,1,17. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Wrong units corrupt every weight. Complete the transfer step: Check dimensions. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Inside weight

Begin with this mechanism: Calculate [1−(u/c)²]² for |u|<c. Recreate the concrete case: Weights are about .66873,.91096,1,.91096. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Weights are nonlinear. Complete the transfer step: Keep digits. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Outside weight

Begin with this mechanism: Set weight zero beyond c. Recreate the concrete case: The value 20 receives zero initially. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Zero weight is model-specific evidence. Complete the transfer step: Investigate point. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Weighted numerator

Begin with this mechanism: Sum w_ix_i. Recreate the concrete case: Keep each contribution visible. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Hidden deletion cannot be audited. Complete the transfer step: Show table. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 9: Weighted denominator

Begin with this mechanism: Sum all positive weights. Recreate the concrete case: The denominator exceeds three. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A tiny denominator signals a failed fit. Complete the transfer step: Monitor it. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: One-step update

Begin with this mechanism: Divide numerator by denominator. Recreate the concrete case: m_1≈2.61684. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One step is not convergence. Complete the transfer step: Continue iterating. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Central root

Begin with this mechanism: Iterate from the robust start. Recreate the concrete case: The symmetric central observations lead to 2.5. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Solver tolerance affects final digits. Complete the transfer step: Record path. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Remote root

Begin with this mechanism: Search alternative starting regions. Recreate the concrete case: m=20 is also a zero-score root with fixed s=1. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Non-convexity creates multiple solutions. Complete the transfer step: Use multi-start. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 13: Efficiency claim

Begin with this mechanism: Tie c≈4.685 to the normal-model setting. Recreate the concrete case: Nominal efficiency is about 95% for the usual M procedure. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Efficiency is model-dependent. Complete the transfer step: Cite source. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Leverage

Begin with this mechanism: In regression inspect x-space leverage. Recreate the concrete case: A small residual can coexist with high leverage. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Residual weighting alone is incomplete. Complete the transfer step: Use robust design method. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Uncertainty

Begin with this mechanism: Choose a covariance or resampling method compatible with the fit. Recreate the concrete case: Report centre and uncertainty together. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Naive least-squares SE can mislead. Complete the transfer step: Validate coverage. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 16: Claim

Begin with this mechanism: State the chosen root, start, scale and c. Recreate the concrete case: The central teaching solution is 2.5. Write observations, starting locations, robust scales, standardized residuals, cutoffs, bisquare scores, IRLS weights, weighted sums, roots and original measurement units with units, signs, thresholds and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call zero-weight observations errors. Complete the transfer step: Keep scope. Change exactly one observation, sampling unit, variance input, confidence level, residual scale, tuning parameter, starting location, score boundary, cluster, leverage value, solver tolerance, resampling rule or covariance formula. Explain which mathematical object changes, which root is selected, which conclusion survives and which claim must weaken. Keep design, standardization, score, weight, objective, solver, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the score table and once to a parent with the real decision and original units. Add a counterexample in which the same robust estimate comes from rounding, censoring, dependence, mixture structure, leverage or genuine rare events. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.