WHY MATHEMATICS?

Why Mathematics? | Martinez–Iglewicz Test: Biweight Scale, Median Absolute Deviation, Long Tails and Robust Normality Evidence

How ordinary variance and bounded biweight spread react differently to extreme observations, and how that contrast becomes a transparent test for long-tailed departures.

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

Choose a reading route

Find the median and median absolute deviation
Convert deviations to bounded biweight coordinates
Calculate robust and ordinary spread estimates
Form the calibrated ratio and inspect influential observations

Chapter map

1. Sample
2. Median
3. Absolute deviations
4. MAD
5. Tuning
6. Biweight u
7. Cutoff
8. Weights
9. Robust numerator
10. Correction denominator
11. Biweight scale
12. Ordinary spread
13. Ratio
14. Critical values
15. Influence plot
16. Claim

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

1. Sample

This is where modelling choices begin to affect evidence. Freeze independent observations.

A concrete situation

Use 1,2,3,4,10.

What the mathematics does not license

Five points are arithmetic only. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Label pedagogy. 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. Median

The central mathematical idea is concrete. Sort and find the middle.

A concrete situation

Median=3.

What the mathematics does not license

Even n needs a declared convention. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

State 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.

3. Absolute deviations

Here the geometry becomes operational. Calculate distances from median.

A concrete situation

2,1,0,1,7.

What the mathematics does not license

Using mean deviations changes the method. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Keep median. 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. MAD

This step deserves a careful pause. Take their median.

A concrete situation

A=1.

What the mathematics does not license

A=0 makes scaling undefined. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

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

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

5. Tuning

The useful question is not only how to calculate, but what the calculation assumes. Use the published constant.

A concrete situation

Set c=9.

What the mathematics does not license

Changing c changes robustness. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Predeclare. 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. Biweight u

A small example makes the mechanism visible. Divide median deviations by cA.

A concrete situation

u values include 7/9.

What the mathematics does not license

Signs matter in some terms. The same reasoning helps students move from an exercise to a real dataset.

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.

7. Cutoff

This is where modelling choices begin to affect evidence. Retain absolute u below one.

A concrete situation

All five remain here.

What the mathematics does not license

Beyond cutoff receives zero weight. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Mark cases. 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. Weights

The central mathematical idea is concrete. Calculate (1-u squared) powers.

A concrete situation

The value 10 is strongly downweighted.

What the mathematics does not license

Weights are not probabilities. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Name them. 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. Robust numerator

Here the geometry becomes operational. Sum weighted squared deviations.

A concrete situation

Central values dominate.

What the mathematics does not license

Rounding early changes high powers. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Keep precision. 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. Correction denominator

This step deserves a careful pause. Sum (1-u squared)(1-5u squared).

A concrete situation

Then square as defined.

What the mathematics does not license

Sign cancellations can destabilize. 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. Biweight scale

The useful question is not only how to calculate, but what the calculation assumes. Apply n and denominator factors.

A concrete situation

Obtain Sb squared.

What the mathematics does not license

Formula variants must not be mixed. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Match 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.

12. Ordinary spread

A small example makes the mechanism visible. Calculate the declared ordinary numerator.

A concrete situation

The tail value contributes strongly.

What the mathematics does not license

Centre convention matters. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Use exact definition. 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. Ratio

This is where modelling choices begin to affect evidence. Divide ordinary spread by corrected biweight scale.

A concrete situation

Near agreement suggests a light-tail fit.

What the mathematics does not license

Near one is not a universal cutoff. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Calibrate by n. 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. Critical values

The central mathematical idea is concrete. Use published simulation or validated resampling.

A concrete situation

Preserve tuning and formula.

What the mathematics does not license

Asymptotic guesses can mislead. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

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

15. Influence plot

Here the geometry becomes operational. Show each ordinary and biweight contribution.

A concrete situation

The value 10 creates contrast.

What the mathematics does not license

Deleting it post hoc biases the test. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Investigate measurement. 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. State robust-scale evidence.

A concrete situation

The sample has unusually long-tail influence.

What the mathematics does not license

Normality and cause are not proven. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

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

A seven-day practice lab

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

Questions for a student, parent or mentor

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

A method-specific audit before deployment

The Martinez–Iglewicz test compares an ordinary spread measure with a robust biweight scale. Under an approximately normal sample, both estimate the same underlying variance and their ratio stays near its calibrated null range. Long tails or influential observations can inflate the ordinary numerator much more than the bounded biweight contribution.

For 1,2,3,4,10, the median is 3 and absolute deviations are 2,1,0,1,7, so the median absolute deviation A is 1. With tuning constant 9, standardized distances u_i=(x_i-3)/9 are -2/9,-1/9,0,1/9,7/9. All remain inside one, but the value 10 receives a much smaller biweight than the central values.

The biweight scale uses squared deviations times (1-u squared) to the fourth power in the numerator and a correction involving (1-u squared)(1-5u squared) in the denominator. Values with absolute u at least one receive zero weight. The formal test statistic is a declared ratio of ordinary and biweight variance estimates, calibrated by simulation or published approximations.

Robust does not mean assumption-free. If more than a few observations come from a different process, the median and MAD can also move. Rounding can make A zero, while skewness can create a large ratio without symmetric heavy tails. Show the weight table, ordinary spread, robust spread and the observations that drive their difference.

Frequently asked questions

What does the Martinez–Iglewicz test compare?

It compares an ordinary variance measure with a robust biweight scale estimate.

Why use the median and MAD?

They provide a resistant centre and initial scale for defining biweight coordinates.

What happens beyond the biweight cutoff?

Observations with absolute standardized distance at least one receive zero biweight contribution.

What alternatives motivated the test?

The paper emphasizes long-tailed symmetric alternatives and also studies skewed cases.

Does robust mean outliers can be ignored?

No. Their provenance and practical magnitude still require investigation.

What if MAD is zero?

The ordinary formulation is undefined and a tie-aware alternative is required.

Primary source and fact-check note

Jorge Martinez and Boris Iglewicz, “A Test for Departure from Normality Based on a Biweight Estimator of Scale,” Biometrika 68(1), 331–333 (1981), DOI 10.1093/biomet/68.1.331. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

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

Extended laboratory: topic-specific validation and responsible use

Design audit

Freeze independent units, target variables and inclusion rules.

Standardization audit

Reproduce centres, scales, covariance roots and projections.

Statistic audit

Show every distance, kernel, moment, direction or weight contribution.

Calibration audit

Match sample size, dimension, tuning, fitting and optimization.

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

Graph audit

Preserve scatterplots, distance plots, direction curves and influence tables.

Magnitude audit

Explain the practical consequence in original units.

Sensitivity audit

Use only prespecified tuning and influence checks.

Final claim

Keep inference inside the model and sampling design.

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

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Sample

Begin with this mechanism: Freeze independent observations. Recreate the concrete case: Use 1,2,3,4,10. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Five points are arithmetic only. Complete the transfer step: Label pedagogy. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 2: Median

Begin with this mechanism: Sort and find the middle. Recreate the concrete case: Median=3. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Even n needs a declared convention. Complete the transfer step: State it. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 3: Absolute deviations

Begin with this mechanism: Calculate distances from median. Recreate the concrete case: 2,1,0,1,7. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Using mean deviations changes the method. Complete the transfer step: Keep median. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 4: MAD

Begin with this mechanism: Take their median. Recreate the concrete case: A=1. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A=0 makes scaling undefined. Complete the transfer step: Check ties. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

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

Workshop 5: Tuning

Begin with this mechanism: Use the published constant. Recreate the concrete case: Set c=9. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Changing c changes robustness. Complete the transfer step: Predeclare. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 6: Biweight u

Begin with this mechanism: Divide median deviations by cA. Recreate the concrete case: u values include 7/9. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Signs matter in some terms. Complete the transfer step: Show table. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 7: Cutoff

Begin with this mechanism: Retain absolute u below one. Recreate the concrete case: All five remain here. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Beyond cutoff receives zero weight. Complete the transfer step: Mark cases. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 8: Weights

Begin with this mechanism: Calculate (1-u squared) powers. Recreate the concrete case: The value 10 is strongly downweighted. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Weights are not probabilities. Complete the transfer step: Name them. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

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

Workshop 9: Robust numerator

Begin with this mechanism: Sum weighted squared deviations. Recreate the concrete case: Central values dominate. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Rounding early changes high powers. Complete the transfer step: Keep precision. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 10: Correction denominator

Begin with this mechanism: Sum (1-u squared)(1-5u squared). Recreate the concrete case: Then square as defined. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Sign cancellations can destabilize. Complete the transfer step: Show terms. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 11: Biweight scale

Begin with this mechanism: Apply n and denominator factors. Recreate the concrete case: Obtain Sb squared. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Formula variants must not be mixed. Complete the transfer step: Match source. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 12: Ordinary spread

Begin with this mechanism: Calculate the declared ordinary numerator. Recreate the concrete case: The tail value contributes strongly. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Centre convention matters. Complete the transfer step: Use exact definition. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

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

Workshop 13: Ratio

Begin with this mechanism: Divide ordinary spread by corrected biweight scale. Recreate the concrete case: Near agreement suggests a light-tail fit. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Near one is not a universal cutoff. Complete the transfer step: Calibrate by n. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 14: Critical values

Begin with this mechanism: Use published simulation or validated resampling. Recreate the concrete case: Preserve tuning and formula. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Asymptotic guesses can mislead. Complete the transfer step: Document. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 15: Influence plot

Begin with this mechanism: Show each ordinary and biweight contribution. Recreate the concrete case: The value 10 creates contrast. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Deleting it post hoc biases the test. Complete the transfer step: Investigate measurement. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.

Workshop 16: Claim

Begin with this mechanism: State robust-scale evidence. Recreate the concrete case: The sample has unusually long-tail influence. Write observations, medians, absolute deviations, MAD values, tuning constants, biweight coordinates, bounded weights, robust numerator terms, correction terms, ordinary spread, scale ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Normality and cause are not proven. Complete the transfer step: Use plots. Change exactly one observation, row, tie, covariance entry, whitening rule, distance, frequency weight, direction-grid resolution, median, MAD, tuning constant, kernel parameter, null approximation or bootstrap seed. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep design, transformation, statistic, calibration, magnitude, interpretation and validation on separate notebook lines. Explain the result twice: first to a student using the visible geometry or weight table, then to a parent using the real decision and original units. Add a counterexample in which the same statistic arises from heavy tails, clusters, curvature, rounding, censoring or dependence. State what extra graph, design record or held-out observation would distinguish the explanations. Finish by writing one sentence the data support, one sentence they do not support and one useful next measurement.