Chapter map

1. 1. Raw series
2. 2. Trend estimator
3. 3. Trend values
4. 4. Detrending
5. 5. Residual mean
6. 6. Lag pairs
7. 7. Lag-one estimate
8. 8. Filtering rule
9. 9. First value
10. 10. Trend restoration
11. 11. Trend test
12. 12. Raw comparison
13. 13. Higher lags
14. 14. Seasonality
15. 15. Simulation
16. 16. Final claim

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

---

1. Raw series

This is where modelling choices begin to affect evidence. Preserve observations and actual timestamps.

A concrete situation

xt is recorded at t=1 through 5.

What the mathematics does not license

Irregular gaps alter lag meaning. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Use real time. 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. Trend estimator

The central mathematical idea is concrete. Choose a slope rule before filtering.

A concrete situation

Illustrative b=1.5 per step.

What the mathematics does not license

Autocorrelation affects slope uncertainty. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Report intervals. 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. Trend values

Here the geometry becomes operational. Calculate bt at every time.

A concrete situation

At t=3 the trend component is 4.5.

What the mathematics does not license

Index origin changes intercept only if consistent. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Declare origin. 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. Detrending

This step deserves a careful pause. Compute yt=xt-bt.

A concrete situation

Residuals are 0,.5,0,-.5,0.

What the mathematics does not license

A nonlinear trend leaves structure. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Plot residuals. 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. Residual mean

The useful question is not only how to calculate, but what the calculation assumes. Centre residuals as required by the estimator.

A concrete situation

The example mean is zero.

What the mathematics does not license

Drift indicates incomplete detrending. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Recheck 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.

---

6. Lag pairs

A small example makes the mechanism visible. Form adjacent residual pairs.

A concrete situation

Four pairs remain.

What the mathematics does not license

Missing points destroy adjacency. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Define gaps. 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. Lag-one estimate

This is where modelling choices begin to affect evidence. Estimate r from detrended residuals.

A concrete situation

Illustrative r=.4.

What the mathematics does not license

Short samples make r unstable. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Show sensitivity. 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. Filtering rule

The central mathematical idea is concrete. Compute y-prime-t=yt-r y(t-1).

A concrete situation

The third residual is -.2.

What the mathematics does not license

Sign mistakes invert dependence. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

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

---

9. First value

Here the geometry becomes operational. Declare how the transform starts.

A concrete situation

The classical series loses the first point.

What the mathematics does not license

Different software may retain it. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Document 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.

---

10. Trend restoration

This step deserves a careful pause. Add bt back to filtered residuals.

A concrete situation

z3=-.2+4.5=4.3.

What the mathematics does not license

Wrong indexing shifts results. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Audit every t. 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. Trend test

The useful question is not only how to calculate, but what the calculation assumes. Apply the selected test to z.

A concrete situation

Mann-Kendall score is recomputed.

What the mathematics does not license

Testing choices cannot be tuned afterward. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

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

---

12. Raw comparison

A small example makes the mechanism visible. Run and plot the raw analysis too.

A concrete situation

Raw and transformed scores are compared.

What the mathematics does not license

Transformation can change power. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Explain differences. 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. Higher lags

This is where modelling choices begin to affect evidence. Inspect residual ACF beyond lag one.

A concrete situation

Lag two remains visible.

What the mathematics does not license

AR(1) filtering is insufficient. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Consider another 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.

---

14. Seasonality

The central mathematical idea is concrete. Remove or model recurring cycles separately.

A concrete situation

Monthly residuals retain seasons.

What the mathematics does not license

TFPW alone does not seasonalize. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Use seasonal 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. Simulation

Here the geometry becomes operational. Generate series with estimated trend and dependence.

A concrete situation

False positives and power are measured.

What the mathematics does not license

Plug-in simulation understates uncertainty. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Vary parameters. 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. Final claim

This step deserves a careful pause. Describe trend evidence under the transformation.

A concrete situation

Results are robust across plausible r.

What the mathematics does not license

Robustness is not causation. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Validate mechanisms. 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

Trend-free prewhitening separates two structures that can otherwise interfere with one another: a persistent trend and short-lag autocorrelation. Classical prewhitening applied directly to the observations can remove part of the trend. The Yue-Pilon procedure first estimates and removes the trend, estimates lag-one dependence from the detrended residuals, filters those residuals and then restores the trend.

Write xt for the observation at time t and b for the estimated trend slope. Detrend with yt=xt-bt. Estimate the lag-one correlation r from the y sequence, form y-prime-t=yt-r y(t-1), and restore the trend as zt=y-prime-t+bt. The transformed z series can then enter the declared trend test.

For an illustrative slope b=1.5 and detrended values 0, 0.5, 0, -0.5, 0, suppose the estimated lag-one correlation is r=0.4. The third filtered residual is 0-0.4(0.5)=-0.2. At t=3, restoring the trend gives z3=-0.2+1.5(3)=4.3. The example shows the sequence of operations; real r and b must be estimated from the actual series.

Trend-free prewhitening is not a universal repair. Short samples make both b and r uncertain, stronger dependence may not be AR(1), and prewhitening can affect variance and power. Responsible work reports the raw and transformed series, slope and correlation estimates, sensitivity choices and simulations that resemble the application.

---

Frequently asked questions

Why detrend before prewhitening?

Filtering the raw series can remove part of the trend together with autocorrelation.

What is prewhitened?

The detrended residual sequence is filtered using its estimated lag-one correlation.

Why restore the trend?

Restoration preserves the estimated long-term component for the subsequent trend analysis.

Is lag-one correlation always sufficient?

No. Higher-order, seasonal or long-memory dependence may require another model.

What happens to the first observation?

The classical transform has no previous residual for it, so edge handling must be declared.

Does transformation guarantee a valid p-value?

No. Estimation uncertainty and the actual dependence structure should be checked with sensitivity analysis and simulation.

---

Primary source and fact-check note

Sheng Yue, Paul Pilon, Bob Phinney and George Cavadias, “The Influence of Autocorrelation on the Ability to Detect Trend in Hydrological Series,” Hydrological Processes 16(9), 1807–1829 (2002), DOI 10.1002/hyp.1095. 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
• Seasonal Kendall trend 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

Time audit

Preserve actual spacing and missingness.

Slope audit

Publish estimator, units and uncertainty.

Detrend audit

Show every residual.

Correlation audit

Estimate r only from detrended structure.

Filter audit

Declare formula and first-value handling.

Restore audit

Reproduce every transformed value.

Simulation audit

Measure size and power under plausible dependence.

Final claim

Describe sensitivity to transformation choices.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Raw series

Begin with this mechanism: Preserve observations and actual timestamps. Recreate the concrete case: xt is recorded at t=1 through 5. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Irregular gaps alter lag meaning. Complete the transfer step: Use real time. 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: Trend estimator

Begin with this mechanism: Choose a slope rule before filtering. Recreate the concrete case: Illustrative b=1.5 per step. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Autocorrelation affects slope uncertainty. Complete the transfer step: Report intervals. 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: Trend values

Begin with this mechanism: Calculate bt at every time. Recreate the concrete case: At t=3 the trend component is 4.5. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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 origin changes intercept only if consistent. Complete the transfer step: Declare origin. 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: Detrending

Begin with this mechanism: Compute yt=xt-bt. Recreate the concrete case: Residuals are 0,.5,0,-.5,0. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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 nonlinear trend leaves structure. Complete the transfer step: Plot residuals. 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: Residual mean

Begin with this mechanism: Centre residuals as required by the estimator. Recreate the concrete case: The example mean is zero. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Drift indicates incomplete detrending. Complete the transfer step: Recheck 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 6: Lag pairs

Begin with this mechanism: Form adjacent residual pairs. Recreate the concrete case: Four pairs remain. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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 points destroy adjacency. Complete the transfer step: Define gaps. 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: Lag-one estimate

Begin with this mechanism: Estimate r from detrended residuals. Recreate the concrete case: Illustrative r=.4. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Short samples make r unstable. Complete the transfer step: Show sensitivity. 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: Filtering rule

Begin with this mechanism: Compute y-prime-t=yt-r y(t-1). Recreate the concrete case: The third residual is -.2. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Sign mistakes invert dependence. Complete the transfer step: Reproduce terms. 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: First value

Begin with this mechanism: Declare how the transform starts. Recreate the concrete case: The classical series loses the first point. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Different software may retain it. Complete the transfer step: Document convention. 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: Trend restoration

Begin with this mechanism: Add bt back to filtered residuals. Recreate the concrete case: z3=-.2+4.5=4.3. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Wrong indexing shifts results. Complete the transfer step: Audit every t. 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: Trend test

Begin with this mechanism: Apply the selected test to z. Recreate the concrete case: Mann-Kendall score is recomputed. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Testing choices cannot be tuned afterward. Complete the transfer step: Pre-register them. 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: Raw comparison

Begin with this mechanism: Run and plot the raw analysis too. Recreate the concrete case: Raw and transformed scores are compared. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Transformation can change power. Complete the transfer step: Explain differences. 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: Higher lags

Begin with this mechanism: Inspect residual ACF beyond lag one. Recreate the concrete case: Lag two remains visible. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: AR(1) filtering is insufficient. Complete the transfer step: Consider another 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 14: Seasonality

Begin with this mechanism: Remove or model recurring cycles separately. Recreate the concrete case: Monthly residuals retain seasons. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: TFPW alone does not seasonalize. Complete the transfer step: Use seasonal 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: Simulation

Begin with this mechanism: Generate series with estimated trend and dependence. Recreate the concrete case: False positives and power are measured. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Plug-in simulation understates uncertainty. Complete the transfer step: Vary parameters. 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: Final claim

Begin with this mechanism: Describe trend evidence under the transformation. Recreate the concrete case: Results are robust across plausible r. Write observations, timestamps, trend slopes, trend values, detrended residuals, lag pairs, lag-one correlations, filtered residuals, restored values, edge conventions, trend scores, simulation sizes and power estimates 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: Robustness is not causation. Complete the transfer step: Validate mechanisms. 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.