Chapter map

1. 1. Sample
2. 2. Order
3. 3. Blom position
4. 4. Normal score
5. 5. Symmetry
6. 6. Weighted products
7. 7. Projection sum
8. 8. Score norm
9. 9. Explained variation
10. 10. Centre data
11. 11. Total variation
12. 12. W-prime
13. 13. Residual fraction
14. 14. Null method
15. 15. Influence
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. Sample

A small example makes the mechanism visible. Freeze independent observations.

A concrete situation

Use 1,2,3,4,8.

What the mathematics does not license

Dependence changes calibration. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

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

---

2. Order

This is where modelling choices begin to affect evidence. Sort observations and preserve IDs.

A concrete situation

x_(i)=1,2,3,4,8.

What the mathematics does not license

A copied rank error changes the projection. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Keep 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.

---

3. Blom position

The central mathematical idea is concrete. Compute (i−3/8)/(n+1/4).

A concrete situation

Use denominator 5.25.

What the mathematics does not license

Using i/(n+1) is another score system. A trustworthy workflow records the choice before seeing the final test result.

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.

---

4. Normal score

Here the geometry becomes operational. Apply Φ⁻¹ to each position.

A concrete situation

Scores are approximately -1.1798,-0.4972,0,0.4972,1.1798.

What the mathematics does not license

Tail rounding changes the squared ratio. Evidence becomes stronger when the limitation is tested instead of hidden.

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.

---

5. Symmetry

This step deserves a careful pause. Check m_i=−m_(n+1−i).

A concrete situation

The centre score is zero.

What the mathematics does not license

Asymmetric scores signal a coding error. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Unit test. 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. Weighted products

The useful question is not only how to calculate, but what the calculation assumes. Multiply m_i by x_(i).

A concrete situation

Large upper observations receive positive tail weight.

What the mathematics does not license

Weights are not probabilities. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Label units. 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. Projection sum

A small example makes the mechanism visible. Add the weighted products.

A concrete situation

Σm_ix_(i)≈9.2527.

What the mathematics does not license

Premature rounding biases the square. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Retain 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.

---

8. Score norm

This is where modelling choices begin to affect evidence. Compute Σm_i².

A concrete situation

The value is about 3.2781.

What the mathematics does not license

Omitting normalization changes W-prime. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

---

9. Explained variation

The central mathematical idea is concrete. Divide the squared projection by the score norm.

A concrete situation

This is the normal-line sum of squares.

What the mathematics does not license

It is geometry, not causal explanation. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use careful language. 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. Centre data

Here the geometry becomes operational. Subtract x̄ from every observation.

A concrete situation

x̄=3.6.

What the mathematics does not license

The projection is translation invariant only with balanced scores. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Verify sum scores. 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. Total variation

This step deserves a careful pause. Compute Σ(x_i−x̄)².

A concrete situation

Total=29.2.

What the mathematics does not license

Using n-divisor variance without multiplying back confuses scale. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Keep sums. 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. W-prime

The useful question is not only how to calculate, but what the calculation assumes. Divide explained by total variation.

A concrete situation

Approximate W-prime≈0.8944.

What the mathematics does not license

It is not a tail probability. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Calibrate. 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. Residual fraction

A small example makes the mechanism visible. Inspect 1−W-prime.

A concrete situation

About 0.1056 is off the normal-score line.

What the mathematics does not license

A single number hides where deviation occurs. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Use Q–Q plot. 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. Null method

This is where modelling choices begin to affect evidence. Choose published transform, table or Monte Carlo.

A concrete situation

Reproduce score construction exactly.

What the mathematics does not license

Mixing calibrations invalidates p-values. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Name version. 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

The central mathematical idea is concrete. Inspect endpoint leverage.

A concrete situation

Eight has the largest score and magnitude.

What the mathematics does not license

High line fit can coexist with important tail shape. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

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

---

16. Claim

Here the geometry becomes operational. State approximate normal-score linearity evidence.

A concrete situation

Ordered observations align well or poorly.

What the mathematics does not license

The statistic does not identify cause. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

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

---

A seven-day practice lab

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

Questions for a student, parent or mentor

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

A method-specific audit before deployment

The Weisberg–Bingham test is a machine-friendly approximation to the Shapiro–Francia W-prime idea. It replaces difficult expected normal order statistics with Blom plotting scores, projects the ordered observations onto that normal-score direction, and divides the explained projection by total centred variation. A value near one means the ordered sample lies close to the straight normal-score pattern.

For 1,2,3,4,8, Blom probabilities (i−3/8)/(n+1/4) give normal scores approximately -1.1798,-0.4972,0,0.4972 and 1.1798. Their weighted sum with the ordered observations is about 9.2527, the sum of squared scores is about 3.2781, and the centred sum of squares is 29.2. The resulting approximate W-prime is about 0.8944.

The ratio has an analysis-of-variance interpretation: the numerator is the squared length of the sample along a normalized normal-score direction, while the denominator is total centred sum of squares. The unexplained fraction 1−W-prime is the part not aligned with the ideal straight line. This geometry is why the statistic is closely related to a probability-plot correlation.

A test still needs the correct null calibration. The statistic depends on sample size, estimated parameters and the exact score formula. Royston-type transformations, tabulated approximations and Monte Carlo procedures are not interchangeable labels for the same p-value. Software output should name the implementation and preserve the score vector so another reader can reproduce the calculation.

---

Frequently asked questions

What are Blom scores?

Standard-normal quantiles at plotting positions (i−3/8)/(n+1/4).

What does the numerator represent?

Squared projection of ordered observations on the normalized normal-score direction.

What does the denominator represent?

Total centred sum of squares in the observations.

What does a value near one mean?

Strong linear alignment with the expected normal-score pattern.

Is W-prime a p-value?

No. It needs a sample-size-matched null calibration.

Why preserve the Q–Q plot?

The scalar ratio cannot show which tail or middle ranks caused misalignment.

---

Primary source and fact-check note

Sanford Weisberg and Christopher Bingham, “An Approximate Analysis of Variance Test for Non-Normality Suitable for Machine Calculation,” Technometrics 17(1), 133–134 (1975), DOI 10.1080/00401706.1975.10489283. 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

• Hosking L-moments for normality
• Filliben probability-plot correlation test
• 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, chronology, target residuals, dependence and inclusion rules.

Transformation audit

Reproduce ordering, plotting positions, normal scores, fitted-null probabilities or standardized moments.

Statistic audit

Show every gap, projection, component, dependence factor, penalty and coefficient.

Calibration audit

Match sample size, parameter estimation, dependence, selection and statistic version.

Graph audit

Preserve Q–Q views, score paths, ACF displays, influence plots and component maps.

Magnitude audit

Explain consequences in original units and decision context.

Sensitivity audit

Use only prespecified validation, model and measurement checks.

Final claim

Keep the conclusion inside the measured population, time window and model.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Sample

Begin with this mechanism: Freeze independent observations. Recreate the concrete case: Use 1,2,3,4,8. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Dependence changes calibration. Complete the transfer step: Audit design. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Order

Begin with this mechanism: Sort observations and preserve IDs. Recreate the concrete case: x_(i)=1,2,3,4,8. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime 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 copied rank error changes the projection. Complete the transfer step: Keep table. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Blom position

Begin with this mechanism: Compute (i−3/8)/(n+1/4). Recreate the concrete case: Use denominator 5.25. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime 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 i/(n+1) is another score system. Complete the transfer step: Match source. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Normal score

Begin with this mechanism: Apply Φ⁻¹ to each position. Recreate the concrete case: Scores are approximately -1.1798,-0.4972,0,0.4972,1.1798. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Tail rounding changes the squared ratio. Complete the transfer step: Keep precision. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Symmetry

Begin with this mechanism: Check m_i=−m_(n+1−i). Recreate the concrete case: The centre score is zero. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Asymmetric scores signal a coding error. Complete the transfer step: Unit test. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Weighted products

Begin with this mechanism: Multiply m_i by x_(i). Recreate the concrete case: Large upper observations receive positive tail weight. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime 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: Label units. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Projection sum

Begin with this mechanism: Add the weighted products. Recreate the concrete case: Σm_ix_(i)≈9.2527. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Premature rounding biases the square. Complete the transfer step: Retain digits. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Score norm

Begin with this mechanism: Compute Σm_i². Recreate the concrete case: The value is about 3.2781. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Omitting normalization changes W-prime. Complete the transfer step: Show denominator. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Explained variation

Begin with this mechanism: Divide the squared projection by the score norm. Recreate the concrete case: This is the normal-line sum of squares. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: It is geometry, not causal explanation. Complete the transfer step: Use careful language. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Centre data

Begin with this mechanism: Subtract x̄ from every observation. Recreate the concrete case: x̄=3.6. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: The projection is translation invariant only with balanced scores. Complete the transfer step: Verify sum scores. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Total variation

Begin with this mechanism: Compute Σ(x_i−x̄)². Recreate the concrete case: Total=29.2. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime 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 n-divisor variance without multiplying back confuses scale. Complete the transfer step: Keep sums. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: W-prime

Begin with this mechanism: Divide explained by total variation. Recreate the concrete case: Approximate W-prime≈0.8944. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: It is not a tail probability. Complete the transfer step: Calibrate. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Residual fraction

Begin with this mechanism: Inspect 1−W-prime. Recreate the concrete case: About 0.1056 is off the normal-score line. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime 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 single number hides where deviation occurs. Complete the transfer step: Use Q–Q plot. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Null method

Begin with this mechanism: Choose published transform, table or Monte Carlo. Recreate the concrete case: Reproduce score construction exactly. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Mixing calibrations invalidates p-values. Complete the transfer step: Name version. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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: Influence

Begin with this mechanism: Inspect endpoint leverage. Recreate the concrete case: Eight has the largest score and magnitude. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: High line fit can coexist with important tail shape. Complete the transfer step: Report plot. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. 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 approximate normal-score linearity evidence. Recreate the concrete case: Ordered observations align well or poorly. Write ordered observations, Blom positions, normal scores, weighted products, projection sums, score norms, total sums of squares, W-prime ratios and calibrated probabilities with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: The statistic does not identify cause. Complete the transfer step: Keep provenance. Change exactly one observation, ordering rule, scale divisor, plotting position, normal score, basis function, nuisance projection, maximum dimension, penalty, lag autocorrelation, dependence factor, rejection direction, finite-sample correction or Monte Carlo 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 once to a student with the component table and once to a parent with the real decision and original units. Add a counterexample in which the same statistic comes from rounding, censoring, mixture structure, dependence, conditional volatility or contamination. State what extra graph or measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.