WHY MATHEMATICS?

Why Mathematics? | Quantile Cointegrating Regression: Non-Stationary Series, Long-Run Quantiles, Fully Modified Estimation and Structural Breaks

How a stable long-run error can coexist with wandering series, why coefficients may vary across innovation quantiles, and why ordinary standard errors fail.

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

Choose a reading route

Diagnose integration and define a plausible long-run relation
Fit and compare cointegrating coefficients across quantiles
Correct endogeneity and nuisance terms with declared procedures
Test breaks, residual stability and forward quantile calibration

Chapter map

1. Series definition
2. Integration order
3. Cointegrating form
4. Predictions
5. Residuals
6. Median loss
7. Alternative slope
8. Alternative loss
9. Quantile coefficient
10. Endogeneity
11. Augmentation
12. Deterministic terms
13. Breaks
14. Inference
15. Forecasting
16. Claim

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 central mathematical idea is concrete. Name each time series and sampling frequency.

A concrete situation

Use x=(1,2,4,5), y=(3,6,9,10).

What the mathematics does not license

Mixed units and revisions distort long-run relations. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Freeze vintages. 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. Integration order

Here the geometry becomes operational. Assess whether levels are non-stationary.

A concrete situation

Treat both as teaching I(1) candidates.

What the mathematics does not license

Four values cannot diagnose a unit root. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use a long sample. 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. Cointegrating form

This step deserves a careful pause. Write y=a+beta x+e.

A concrete situation

Set a=1,beta=2.

What the mathematics does not license

A regression equation alone is not cointegration. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Inspect e over 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.

4. Predictions

The useful question is not only how to calculate, but what the calculation assumes. Evaluate 1+2x.

A concrete situation

Get 3,5,9,11.

What the mathematics does not license

One intercept convention changes residuals. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

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

5. Residuals

A small example makes the mechanism visible. Subtract predictions from y.

A concrete situation

Get 0,1,0,-1.

What the mathematics does not license

Bounded values are illustrative, not a stationarity test. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

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

6. Median loss

This is where modelling choices begin to affect evidence. Sum half absolute residuals.

A concrete situation

Loss is 1.

What the mathematics does not license

Serial dependence changes uncertainty. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Use valid inference. 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. Alternative slope

The central mathematical idea is concrete. Try beta=1.5 with a=1.

A concrete situation

Predictions are 2.5,4,7,8.5.

What the mathematics does not license

Candidate comparison is not global proof. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Solve full objective. 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. Alternative loss

Here the geometry becomes operational. Use residuals .5,2,2,1.5.

A concrete situation

Median loss is 3.

What the mathematics does not license

Training preference can fail later. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Validate forward. 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. Quantile coefficient

This step deserves a careful pause. Repeat over tau.

A concrete situation

Let long-run slope vary by innovation quantile.

What the mathematics does not license

Crossing and weak tail support matter. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Plot paths. 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. Endogeneity

The useful question is not only how to calculate, but what the calculation assumes. Allow x innovations to correlate with e.

A concrete situation

Use fully modified correction.

What the mathematics does not license

Correction needs long-run covariance estimates. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Report bandwidth. 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. Augmentation

A small example makes the mechanism visible. Add leads or lags of differences when justified.

A concrete situation

Absorb short-run correlation.

What the mathematics does not license

Over-augmentation wastes information. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Select transparently. 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. Deterministic terms

This is where modelling choices begin to affect evidence. Choose intercept, trend and seasonality.

A concrete situation

Compare a trend specification.

What the mathematics does not license

Omitted trend can fake cointegration. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Run 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.

13. Breaks

The central mathematical idea is concrete. Test known or plausible regime changes.

A concrete situation

Split before and after a policy shift.

What the mathematics does not license

Searching every break overfits. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Predeclare logic. 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. Inference

Here the geometry becomes operational. Use cointegration-specific limiting theory.

A concrete situation

Construct Wald tests from corrected estimates.

What the mathematics does not license

IID t tests are invalid. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Match theory. 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. Forecasting

This step deserves a careful pause. Evaluate future quantile hits and loss.

A concrete situation

Use rolling origins.

What the mathematics does not license

Long-run fit need not predict short horizons. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

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

16. Claim

The useful question is not only how to calculate, but what the calculation assumes. State sample, tau, integration and correction.

A concrete situation

Teaching losses are 1 and 3.

What the mathematics does not license

Do not call correlation equilibrium or causation. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

A seven-day practice lab

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

Questions for a student, parent or mentor

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

A method-specific audit before deployment

Quantile cointegrating regression studies long-run relationships between non-stationary series when the equilibrium coefficient may differ across innovation quantiles. Two trending series can look related simply because both wander over time, so ordinary regression rules do not apply automatically. Xiao derived quantile-regression inference for cointegrated series, proposed quantile-varying cointegrating coefficients, and developed fully modified and augmented procedures to address endogeneity, second-order bias and nuisance parameters.

Take x=(1,2,4,5) and y=(3,6,9,10). Under y=1+2x+e, residuals are (0,1,0,-1), bounded around zero, and total median check loss is 1. With intercept 1 and slope 1.5, predictions are (2.5,4,7,8.5), residuals are (.5,2,2,1.5), and median check loss is 3. The slope-two relation fits this teaching sample better, but four observations cannot establish unit roots, cointegration or a population quantile coefficient.

Cointegration is about stochastic order and a stable long-run error, not merely a high correlation or a visually parallel chart. Structural breaks can make a formerly stable residual wander. Quantile-varying coefficients may reflect asymmetric adjustment, changing volatility or model misspecification. Endogenous integrated regressors change limiting distributions, so conventional IID standard errors and random train-test splits are inappropriate.

Report the variables, units, transformations, sampling frequency, integration evidence, deterministic terms, break treatment, tau grid, lag augmentation, fully modified corrections, bandwidths, solver and time-respecting validation. Distinguish a long-run conditional quantile relation from short-run forecasting and from causation. Date the sample and test stability beyond the estimation period.

Frequently asked questions

What is cointegration?

A combination of non-stationary series has a stable lower-order error process under the model.

Why use quantile regression?

The long-run coefficient and adjustment may differ across innovation quantiles.

Why not use ordinary t tests?

Integrated regressors create nonstandard limiting distributions and endogeneity terms.

What does fully modified estimation address?

Second-order bias and nuisance effects from long-run dependence and endogeneity.

Why test structural breaks?

A changing equilibrium can make a full-sample residual look unstable.

What must be reported?

Variables, integration evidence, deterministic terms, tau, corrections, breaks and forward validation.

Primary source and fact-check note

Zhijie Xiao, “Quantile Cointegrating Regression,” Journal of Econometrics 150(2), 248–260 (2009), DOI 10.1016/j.jeconom.2008.12.005. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

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

Extended laboratory: topic-specific validation and responsible use

Design audit

Freeze time-series vintages, sampling frequency, margins, pseudo-observations, panel dimensions, missing-data rules, running variable, cutoff and treatment assignment before estimation.

Objective audit

Write the cointegrating check loss, copula conditional-CDF plug-in, low-rank quantile-factor objective or local distribution-regression discontinuity estimand explicitly.

Tuning audit

Record tau grids, lag augmentation, long-run covariance bandwidth, copula family, vine order, factor count, starts, RD bandwidth, kernel, polynomial order and CDF grid.

Computation audit

Check residual integration, correction terms, response weights, copula likelihood, alternating-factor convergence, local CDF monotonicity, first-stage jumps and numerical tolerance.

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

Support audit

Inspect time breaks, marginal and joint tails, concentrated copula weights, sparse panel cells, quantile-specific factor support and observations immediately around the cutoff.

Uncertainty audit

Propagate long-run dependence, marginal estimation, copula fitting, factor selection, nonconvex optimisation, local bandwidth and CDF inversion uncertainty through the complete workflow.

Validation audit

Use future periods, held-out units, blocked panels or bandwidth-stable cutoff windows for quantile hits, weighted-CDF calibration, reconstruction loss and distributional effect sensitivity.

Final claim

Translate the result without turning cointegration, copula dependence, a latent factor or a local discontinuity effect into a universal causal guarantee.

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

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Series definition

Begin with this mechanism: Name each time series and sampling frequency. Recreate the concrete case: Use x=(1,2,4,5), y=(3,6,9,10). Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Mixed units and revisions distort long-run relations. Complete the transfer step: Freeze vintages. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Integration order

Begin with this mechanism: Assess whether levels are non-stationary. Recreate the concrete case: Treat both as teaching I(1) candidates. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Four values cannot diagnose a unit root. Complete the transfer step: Use a long sample. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Cointegrating form

Begin with this mechanism: Write y=a+beta x+e. Recreate the concrete case: Set a=1,beta=2. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A regression equation alone is not cointegration. Complete the transfer step: Inspect e over time. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Predictions

Begin with this mechanism: Evaluate 1+2x. Recreate the concrete case: Get 3,5,9,11. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One intercept convention changes residuals. Complete the transfer step: Show table. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

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

Workshop 5: Residuals

Begin with this mechanism: Subtract predictions from y. Recreate the concrete case: Get 0,1,0,-1. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Bounded values are illustrative, not a stationarity test. Complete the transfer step: Plot and test. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Median loss

Begin with this mechanism: Sum half absolute residuals. Recreate the concrete case: Loss is 1. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Serial dependence changes uncertainty. Complete the transfer step: Use valid inference. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Alternative slope

Begin with this mechanism: Try beta=1.5 with a=1. Recreate the concrete case: Predictions are 2.5,4,7,8.5. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Candidate comparison is not global proof. Complete the transfer step: Solve full objective. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Alternative loss

Begin with this mechanism: Use residuals .5,2,2,1.5. Recreate the concrete case: Median loss is 3. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Training preference can fail later. Complete the transfer step: Validate forward. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

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

Workshop 9: Quantile coefficient

Begin with this mechanism: Repeat over tau. Recreate the concrete case: Let long-run slope vary by innovation quantile. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Crossing and weak tail support matter. Complete the transfer step: Plot paths. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Endogeneity

Begin with this mechanism: Allow x innovations to correlate with e. Recreate the concrete case: Use fully modified correction. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Correction needs long-run covariance estimates. Complete the transfer step: Report bandwidth. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Augmentation

Begin with this mechanism: Add leads or lags of differences when justified. Recreate the concrete case: Absorb short-run correlation. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Over-augmentation wastes information. Complete the transfer step: Select transparently. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Deterministic terms

Begin with this mechanism: Choose intercept, trend and seasonality. Recreate the concrete case: Compare a trend specification. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Omitted trend can fake cointegration. Complete the transfer step: Run sensitivity. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

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

Workshop 13: Breaks

Begin with this mechanism: Test known or plausible regime changes. Recreate the concrete case: Split before and after a policy shift. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Searching every break overfits. Complete the transfer step: Predeclare logic. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Inference

Begin with this mechanism: Use cointegration-specific limiting theory. Recreate the concrete case: Construct Wald tests from corrected estimates. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: IID t tests are invalid. Complete the transfer step: Match theory. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Forecasting

Begin with this mechanism: Evaluate future quantile hits and loss. Recreate the concrete case: Use rolling origins. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Long-run fit need not predict short horizons. Complete the transfer step: Separate goals. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 16: Claim

Begin with this mechanism: State sample, tau, integration and correction. Recreate the concrete case: Teaching losses are 1 and 3. Write time indexes, original series units, integration orders, intercepts, long-run slopes, residuals, quantile levels, check losses, correction terms, bandwidths and forecast horizons with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call correlation equilibrium or causation. Complete the transfer step: Bound conclusion. Change exactly one observation, series transformation, long-run slope, residual, tau level, lag, correction bandwidth, marginal rank, copula family, query weight, response value, factor count, loading, factor path, missing panel cell, running-variable value, cutoff, bandwidth, kernel weight, treatment jump, CDF threshold or validation block. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, tuning, computation, support, uncertainty and interpretation on separate notebook lines. Explain the result once to a student with the complete table and once to a parent with the real decision and original units. Add a counterexample involving spurious trends, structural breaks, misspecified margins, tail-dependence failure, concentrated weights, factor rotation, local optima, cutoff manipulation, weak first stages or simultaneous policy changes. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.