Chapter map

1. 1. Series definition
2. 2. Full mean
3. 3. Centred values
4. 4. First partial sum
5. 5. Second partial sum
6. 6. Third partial sum
7. 7. Return path
8. 8. Sample scale
9. 9. Standardized path
10. 10. Range R
11. 11. Candidate location
12. 12. Null simulation
13. 13. Dependence check
14. 14. Seasonality check
15. 15. Metadata review
16. 16. Transfer validation

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

---

1. Series definition

The useful question is not only how to calculate, but what the calculation assumes. Fix ordered observations and time units.

A concrete situation

The series is 2,3,4,9,10.

What the mathematics does not license

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

Try it, check it, transfer it

Predefine dates. 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. Full mean

A small example makes the mechanism visible. Compute X-bar once.

A concrete situation

Mean=5.6.

What the mathematics does not license

Outliers influence the centre. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Report robust checks. 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. Centred values

This is where modelling choices begin to affect evidence. Subtract the mean from every point.

A concrete situation

First deviation is -3.6.

What the mathematics does not license

Rounding breaks the zero-sum check. The habit of checking assumptions is one of the practical benefits of learning mathematics.

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.

---

4. First partial sum

The central mathematical idea is concrete. Start S-star at zero and accumulate.

A concrete situation

S1=-3.6.

What the mathematics does not license

Index errors move the path. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Label k clearly. 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. Second partial sum

Here the geometry becomes operational. Add the next deviation.

A concrete situation

S2=-6.2.

What the mathematics does not license

A run can arise by chance. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Use null calibration. 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. Third partial sum

This step deserves a careful pause. Continue the same recursion.

A concrete situation

S3=-7.8.

What the mathematics does not license

The minimum is sample-dependent. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Record all values. 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. Return path

The useful question is not only how to calculate, but what the calculation assumes. Add positive later deviations.

A concrete situation

S4=-4.4 and S5=0.

What the mathematics does not license

Failure to return signals arithmetic error. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Check the sum. 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. Sample scale

A small example makes the mechanism visible. Calculate s with a stated divisor.

A concrete situation

Sample s≈3.647.

What the mathematics does not license

Population and sample definitions differ. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

State the convention. 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. Standardized path

This is where modelling choices begin to affect evidence. Divide every partial sum by s.

A concrete situation

Minimum≈-2.139.

What the mathematics does not license

Scale estimation is uncertain. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Use correct critical values. 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. Range R

The central mathematical idea is concrete. Subtract minimum from maximum.

A concrete situation

R≈2.139 because max is zero.

What the mathematics does not license

R is not a p-value. The same reasoning helps students move from an exercise to a real dataset.

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.

---

11. Candidate location

Here the geometry becomes operational. Inspect extrema and path shape.

A concrete situation

Minimum occurs at k=3.

What the mathematics does not license

A broad valley gives uncertain timing. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Show the 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.

---

12. Null simulation

This step deserves a careful pause. Generate homogeneous independent series.

A concrete situation

Each replicate recomputes mean, s and range.

What the mathematics does not license

Normal simulation may be wrong. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Match the data 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.

---

13. Dependence check

The useful question is not only how to calculate, but what the calculation assumes. Measure serial structure.

A concrete situation

Positive autocorrelation lengthens excursions.

What the mathematics does not license

Independent limits become liberal. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Use block simulation. 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. Seasonality check

A small example makes the mechanism visible. Remove or model recurring cycles.

A concrete situation

Wet and dry seasons differ.

What the mathematics does not license

A cycle can mimic a block shift. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Compare like periods. 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. Metadata review

This is where modelling choices begin to affect evidence. Check instruments and procedures near candidates.

A concrete situation

A gauge relocation is documented.

What the mathematics does not license

Coincidence is not proof. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Keep records separate. 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. Transfer validation

The central mathematical idea is concrete. Apply frozen rules to another series or period.

A concrete situation

A neighbouring record is checked.

What the mathematics does not license

Shared regional change complicates attribution. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Use several references. 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 Buishand range test studies whether a sequence can reasonably be treated as homogeneous in mean. It subtracts the full-series average from each observation, accumulates those deviations and measures how widely the cumulative path travels. A sustained block below the overall mean drives the path downward; a later block above the mean brings it back toward zero.

For observations X1,…,Xn with mean X̄, define S*k=Σi≤k(Xi−X̄), with S*0=S*n=0. Standardize by the sample standard deviation s and calculate the range R=max(S*k/s)−min(S*k/s). Critical values or Monte Carlo simulation calibrate the complete path under the homogeneous null.

For 2, 3, 4, 9 and 10, the mean is 5.6 and deviations are −3.6, −2.6, −1.6, 3.4 and 4.4. The cumulative deviations are −3.6, −6.2, −7.8, −4.4 and 0. With sample s≈3.647, the standardized range is 7.8/3.647≈2.139. That number needs a sample-size-specific null before interpretation.

The test is sensitive to assumptions about independence, distribution, seasonality and the selected period. A large range identifies cumulative inconsistency, not its cause or a guaranteed unique change time. Responsible work plots the path, reports scale and calibration, checks dependence and validates any suspected break against records.

---

Frequently asked questions

What does the Buishand range test assess?

It assesses whether cumulative deviations are consistent with a homogeneous mean under the declared null.

Why do cumulative deviations end at zero?

Every observation is centred by the full-series mean, so all centred values sum to zero.

What is the range statistic?

It is the difference between the maximum and minimum standardized cumulative deviations.

Does the largest excursion prove a change date?

No. It provides evidence of inhomogeneity and a location clue that needs calibrated inference and context.

Why standardize by s?

It places cumulative movement on the scale of ordinary variation so series with different units can be evaluated.

Can the test handle autocorrelation automatically?

No. Dependence changes the null path and should be modelled or preserved in calibration.

---

Primary source and fact-check note

T. A. Buishand, “Some Methods for Testing the Homogeneity of Rainfall Records,” Journal of Hydrology 58(1–2), 11–27 (1982), DOI 10.1016/0022-1694(82)90066-X. 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

• Pettitt change-point test
• Page–Hinkley change detection
• 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

Window audit

Predefine the analysed period.

Arithmetic audit

Verify mean, deviations and zero-sum path.

Scale audit

State sample-standard-deviation convention.

Range audit

Publish every cumulative value and extrema.

Null audit

Calibrate the complete path under realistic dependence.

Season audit

Remove or model recurring cycles.

Metadata audit

Check candidate breaks against records.

Final claim

Report inhomogeneity evidence, not automatic cause.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Series definition

Begin with this mechanism: Fix ordered observations and time units. Recreate the concrete case: The series is 2,3,4,9,10. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Changing the window changes the mean. Complete the transfer step: Predefine dates. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 2: Full mean

Begin with this mechanism: Compute X-bar once. Recreate the concrete case: Mean=5.6. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Outliers influence the centre. Complete the transfer step: Report robust checks. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 3: Centred values

Begin with this mechanism: Subtract the mean from every point. Recreate the concrete case: First deviation is -3.6. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Rounding breaks the zero-sum check. Complete the transfer step: Keep precision. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 4: First partial sum

Begin with this mechanism: Start S-star at zero and accumulate. Recreate the concrete case: S1=-3.6. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Index errors move the path. Complete the transfer step: Label k clearly. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 5: Second partial sum

Begin with this mechanism: Add the next deviation. Recreate the concrete case: S2=-6.2. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: A run can arise by chance. Complete the transfer step: Use null calibration. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 6: Third partial sum

Begin with this mechanism: Continue the same recursion. Recreate the concrete case: S3=-7.8. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: The minimum is sample-dependent. Complete the transfer step: Record all values. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 7: Return path

Begin with this mechanism: Add positive later deviations. Recreate the concrete case: S4=-4.4 and S5=0. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Failure to return signals arithmetic error. Complete the transfer step: Check the sum. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 8: Sample scale

Begin with this mechanism: Calculate s with a stated divisor. Recreate the concrete case: Sample s≈3.647. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Population and sample definitions differ. Complete the transfer step: State the convention. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 9: Standardized path

Begin with this mechanism: Divide every partial sum by s. Recreate the concrete case: Minimum≈-2.139. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Scale estimation is uncertain. Complete the transfer step: Use correct critical values. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 10: Range R

Begin with this mechanism: Subtract minimum from maximum. Recreate the concrete case: R≈2.139 because max is zero. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: R is not a p-value. Complete the transfer step: Calibrate by n. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 11: Candidate location

Begin with this mechanism: Inspect extrema and path shape. Recreate the concrete case: Minimum occurs at k=3. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: A broad valley gives uncertain timing. Complete the transfer step: Show the plot. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 12: Null simulation

Begin with this mechanism: Generate homogeneous independent series. Recreate the concrete case: Each replicate recomputes mean, s and range. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Normal simulation may be wrong. Complete the transfer step: Match the data model. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 13: Dependence check

Begin with this mechanism: Measure serial structure. Recreate the concrete case: Positive autocorrelation lengthens excursions. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Independent limits become liberal. Complete the transfer step: Use block simulation. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 14: Seasonality check

Begin with this mechanism: Remove or model recurring cycles. Recreate the concrete case: Wet and dry seasons differ. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: A cycle can mimic a block shift. Complete the transfer step: Compare like periods. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 15: Metadata review

Begin with this mechanism: Check instruments and procedures near candidates. Recreate the concrete case: A gauge relocation is documented. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Coincidence is not proof. Complete the transfer step: Keep records separate. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.

Workshop 16: Transfer validation

Begin with this mechanism: Apply frozen rules to another series or period. Recreate the concrete case: A neighbouring record is checked. Write observations, dates, full-series means, centred deviations, cumulative sums, sample standard deviations, standardized paths, maxima, minima, ranges, simulation maxima and metadata events with units, denominators and assumptions. Calculate the smallest non-trivial example by hand and reproduce it in a spreadsheet or short program. Challenge this limitation: Shared regional change complicates attribution. Complete the transfer step: Use several references. Change exactly one observation, date, analysis window, reference station, reference weight, mean, standard deviation, split point, segment length, cumulative deviation, range, tie rule, pair sign, cut-off length, probability level, critical value, autocorrelation estimate, null model, simulation seed, seasonal grouping, metadata event or validation period. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, complete-search calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods or reference series outside tuning when testing transfer.