Chapter map

1. 1. Season definition
2. 2. Time alignment
3. 3. Available pairs
4. 4. Sign calculation
5. 5. Seasonal score
6. 6. Second season
7. 7. Overall score
8. 8. Tie groups
9. 9. Season variance
10. 10. Cross covariance
11. 11. Standardization
12. 12. Seasonal slope
13. 13. Flow adjustment
14. 14. Serial dependence
15. 15. Changing seasonality
16. 16. Attribution

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

---

1. Season definition

A small example makes the mechanism visible. Choose recurring groups with substantive meaning.

A concrete situation

Months form twelve seasons.

What the mathematics does not license

Data-driven seasons bias inference. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Freeze labels. 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. Time alignment

This is where modelling choices begin to affect evidence. Align each season across years.

A concrete situation

January values share one ordered series.

What the mathematics does not license

Calendar changes break comparability. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Audit timestamps. 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. Available pairs

The central mathematical idea is concrete. List i<j within each season.

A concrete situation

Three years give three pairs per season.

What the mathematics does not license

Missing years reduce information. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

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

---

4. Sign calculation

Here the geometry becomes operational. Use +1, -1 or 0 for later-minus-earlier.

A concrete situation

Season A gives +1,+1,+1.

What the mathematics does not license

Direction errors reverse the claim. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

State orientation. 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. Seasonal score

This step deserves a careful pause. Add signs inside a season.

A concrete situation

SA=3.

What the mathematics does not license

A score depends on pair count. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

---

6. Second season

The useful question is not only how to calculate, but what the calculation assumes. Repeat without cross-season pairs.

A concrete situation

Season B gives -1,+1,+1.

What the mathematics does not license

One reversal can matter in small samples. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show every sign. 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. Overall score

A small example makes the mechanism visible. Sum declared seasonal scores.

A concrete situation

S=3+1=4.

What the mathematics does not license

The total is not yet a probability. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Calculate its null. 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. Tie groups

This is where modelling choices begin to affect evidence. Record repeated values by season.

A concrete situation

Rounded measurements create ties.

What the mathematics does not license

Ignoring ties inflates information. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Use tie corrections. 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. Season variance

The central mathematical idea is concrete. Compute the null variance for each seasonal score.

A concrete situation

Each three-point no-tie season has its variance.

What the mathematics does not license

Tiny samples defeat approximations. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use exact methods when feasible. 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. Cross covariance

Here the geometry becomes operational. Include covariance under the chosen multivariate version.

A concrete situation

Adjacent monthly seasons may co-move.

What the mathematics does not license

Assuming zero covariance can overstate precision. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Declare the formula. 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. Standardization

This step deserves a careful pause. Divide the corrected score by its standard error.

A concrete situation

A Z statistic summarizes evidence.

What the mathematics does not license

Normal approximation may be poor. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Check 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.

---

12. Seasonal slope

The useful question is not only how to calculate, but what the calculation assumes. Estimate magnitude from same-season slopes.

A concrete situation

January-to-January slopes have annual units.

What the mathematics does not license

Pooling unequal gaps carelessly changes units. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use actual times. 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. Flow adjustment

A small example makes the mechanism visible. Separate known drivers where scientifically justified.

A concrete situation

Concentration residuals may be flow-adjusted.

What the mathematics does not license

Bad adjustment creates artefacts. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Validate the 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.

---

14. Serial dependence

This is where modelling choices begin to affect evidence. Inspect lag structure within seasons.

A concrete situation

Persistent annual residuals remain.

What the mathematics does not license

The basic null is not exact. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Use a dependence-aware 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.

---

15. Changing seasonality

The central mathematical idea is concrete. Test whether seasonal shape itself changes.

A concrete situation

Wet-season timing shifts.

What the mathematics does not license

Fixed seasons can hide the mechanism. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Plot every season. 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. Attribution

Here the geometry becomes operational. Compare trend evidence with independent records.

A concrete situation

A treatment change occurs nearby.

What the mathematics does not license

Coincidence is not cause. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Separate detection and explanation. 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 seasonal Kendall test keeps recurring seasons separate while asking whether values within each season tend to rise or fall over years. January is compared with January, quarter one with quarter one, or the dry season with the dry season. That design prevents a large, predictable seasonal cycle from being mistaken for a long-term trend.

For each season, compare every later observation with every earlier observation and add +1, -1 or 0. The seasonal score is the sum of those signs, and the overall score is the sum across seasons. Its variance must include ties and, where seasons are correlated, the covariance terms appropriate to the selected version.

Suppose season A has values 2, 3 and 5 over three years. Its three pair signs are +1,+1,+1, so SA=3. Season B has 10, 9 and 11; its signs are -1,+1,+1, so SB=1. The total S is 4. No A value is compared with a B value, which is the mathematical protection against confusing season with trend.

The test does not automatically solve serial correlation, changing seasonality, irregular sampling, detection limits or flow effects. A result supports a monotonic tendency within the declared seasons. It does not prove a constant slope, a causal mechanism or the absence of important reversals.

---

Frequently asked questions

What does the seasonal Kendall test assess?

It assesses whether observations tend to move monotonically over time within the same recurring seasons.

Why avoid comparisons across seasons?

Cross-season pairs can confuse a predictable seasonal level difference with a long-term trend.

How are seasonal scores combined?

Pairwise sign scores are calculated within each season and then summed, with an appropriate variance and covariance calculation.

Can it handle missing months?

It can use available within-season pairs, but missingness mechanisms and changing coverage still need investigation.

Does it automatically solve serial correlation?

No. Serial and cross-season dependence can change the null variance and should be handled explicitly.

Does significance imply a constant yearly slope?

No. It supports monotonic ordering, while a separate seasonal slope estimator measures magnitude.

---

Primary source and fact-check note

Robert M. Hirsch, James R. Slack and Richard A. Smith, “Techniques of Trend Analysis for Monthly Water Quality Data,” Water Resources Research 18(1), 107–121 (1982), DOI 10.1029/WR018i001p00107. 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

• Mann–Kendall trend test
• Buishand range 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

Season audit

Justify recurring groups before testing.

Pair audit

Reproduce every within-season sign.

Tie audit

Preserve measurement resolution.

Variance audit

Include correct seasonal variance and covariance.

Dependence audit

Inspect serial and cross-season persistence.

Slope audit

Report a separate seasonal effect size.

Mechanism audit

Check flow, procedure and coverage changes.

Final claim

Describe within-season monotonic evidence.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Season definition

Begin with this mechanism: Choose recurring groups with substantive meaning. Recreate the concrete case: Months form twelve seasons. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Data-driven seasons bias inference. Complete the transfer step: Freeze labels. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 2: Time alignment

Begin with this mechanism: Align each season across years. Recreate the concrete case: January values share one ordered series. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Calendar changes break comparability. Complete the transfer step: Audit timestamps. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 3: Available pairs

Begin with this mechanism: List i<j within each season. Recreate the concrete case: Three years give three pairs per season. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Missing years reduce information. Complete the transfer step: Count pairs. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 4: Sign calculation

Begin with this mechanism: Use +1, -1 or 0 for later-minus-earlier. Recreate the concrete case: Season A gives +1,+1,+1. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Direction errors reverse the claim. Complete the transfer step: State orientation. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 5: Seasonal score

Begin with this mechanism: Add signs inside a season. Recreate the concrete case: SA=3. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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 score depends on pair count. Complete the transfer step: Report n and tau. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 6: Second season

Begin with this mechanism: Repeat without cross-season pairs. Recreate the concrete case: Season B gives -1,+1,+1. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: One reversal can matter in small samples. Complete the transfer step: Show every sign. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 7: Overall score

Begin with this mechanism: Sum declared seasonal scores. Recreate the concrete case: S=3+1=4. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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 total is not yet a probability. Complete the transfer step: Calculate its null. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 8: Tie groups

Begin with this mechanism: Record repeated values by season. Recreate the concrete case: Rounded measurements create ties. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Ignoring ties inflates information. Complete the transfer step: Use tie corrections. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 9: Season variance

Begin with this mechanism: Compute the null variance for each seasonal score. Recreate the concrete case: Each three-point no-tie season has its variance. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Tiny samples defeat approximations. Complete the transfer step: Use exact methods when feasible. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 10: Cross covariance

Begin with this mechanism: Include covariance under the chosen multivariate version. Recreate the concrete case: Adjacent monthly seasons may co-move. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Assuming zero covariance can overstate precision. Complete the transfer step: Declare the formula. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 11: Standardization

Begin with this mechanism: Divide the corrected score by its standard error. Recreate the concrete case: A Z statistic summarizes evidence. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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 approximation may be poor. Complete the transfer step: Check simulation. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 12: Seasonal slope

Begin with this mechanism: Estimate magnitude from same-season slopes. Recreate the concrete case: January-to-January slopes have annual units. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Pooling unequal gaps carelessly changes units. Complete the transfer step: Use actual times. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 13: Flow adjustment

Begin with this mechanism: Separate known drivers where scientifically justified. Recreate the concrete case: Concentration residuals may be flow-adjusted. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Bad adjustment creates artefacts. Complete the transfer step: Validate the model. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 14: Serial dependence

Begin with this mechanism: Inspect lag structure within seasons. Recreate the concrete case: Persistent annual residuals remain. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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 basic null is not exact. Complete the transfer step: Use a dependence-aware design. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 15: Changing seasonality

Begin with this mechanism: Test whether seasonal shape itself changes. Recreate the concrete case: Wet-season timing shifts. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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: Fixed seasons can hide the mechanism. Complete the transfer step: Plot every season. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.

Workshop 16: Attribution

Begin with this mechanism: Compare trend evidence with independent records. Recreate the concrete case: A treatment change occurs nearby. Write seasons, years, timestamps, observations, available within-season pairs, positive signs, negative signs, ties, seasonal scores, total scores, variances, covariances, standardized statistics and seasonal slopes 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 cause. Complete the transfer step: Separate detection and explanation. Change exactly one observation, timestamp, season label, missing-value rule, tie rule, slope estimate, autocorrelation estimate, retained lag, rank, pair sign, variance formula, covariance term, filter convention, null distribution, simulation seed, reference period or validation window. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, preprocessing, statistic, calibration, uncertainty, interpretation and independent validation on separate notebook lines. Reserve complete periods outside tuning when testing transfer.