WHY MATHEMATICS?

Why Mathematics? | Quantile Regression for Longitudinal Data: Individual Effects, L1 Penalties, Repeated Measures and Clustered Uncertainty

How repeated observations separate stable unit levels from common quantile slopes, why shrinkage controls many fixed effects, and how dependence changes validation.

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

Define panel units, time and the target conditional quantile
Separate individual effects from common covariate slopes
Regularise high-dimensional unit effects with an explicit penalty
Handle within-unit dependence, missingness and cluster-aware uncertainty

Chapter map

1. Panel unit
2. Time
3. Outcomes
4. Target
5. Individual effects
6. Common slope
7. Residuals
8. Pooled model
9. Pooled residuals
10. Pooled loss
11. Penalty
12. Identification
13. Unbalanced panels
14. Missingness
15. Uncertainty
16. Claim

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

1. Panel unit

This step deserves a careful pause. Name the repeated entity.

A concrete situation

Use students A and B.

What the mathematics does not license

Rows within a unit are dependent. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Keep IDs. 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

The useful question is not only how to calculate, but what the calculation assumes. Order visits 0 and 1.

A concrete situation

Observe each student twice.

What the mathematics does not license

Two visits provide little within-unit information. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record timing. 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. Outcomes

A small example makes the mechanism visible. Use A:2,3 and B:5,6.

A concrete situation

Student B has a higher stable level.

What the mathematics does not license

Values are illustrative only. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

4. Target

This is where modelling choices begin to affect evidence. Choose tau=.5.

A concrete situation

Fit a conditional median.

What the mathematics does not license

Another tau can give another slope. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use grid later. 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. Individual effects

The central mathematical idea is concrete. Set alpha_A=2 and alpha_B=5.

A concrete situation

Absorb baseline difference.

What the mathematics does not license

One effect per unit increases dimension. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Normalise. 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. Common slope

Here the geometry becomes operational. Set beta=1 for time.

A concrete situation

Predict both second outcomes.

What the mathematics does not license

A time trend need not be causal. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Keep interpretation. 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. Residuals

This step deserves a careful pause. Subtract fitted values.

A concrete situation

All four residuals are zero.

What the mathematics does not license

Perfect fit can signal saturation. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Check degrees of freedom. 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. Pooled model

The useful question is not only how to calculate, but what the calculation assumes. Use intercept 3.5 and slope 1.

A concrete situation

Ignore unit levels.

What the mathematics does not license

Pooling mixes within and between variation. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Compare explicitly. 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. Pooled residuals

A small example makes the mechanism visible. Get -1.5,-1.5,1.5,1.5 in unit-time order.

A concrete situation

Absolute residual sum is 6.

What the mathematics does not license

Order must remain auditable. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

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

10. Pooled loss

This is where modelling choices begin to affect evidence. Multiply by .5 at the median.

A concrete situation

Total check loss is 3.

What the mathematics does not license

Training loss alone cannot choose a model. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Validate. 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. Penalty

The central mathematical idea is concrete. Shrink individual effects under a stated parameterisation.

A concrete situation

Tune lambda.

What the mathematics does not license

Penalty changes the estimand and bias. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Publish 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.

12. Identification

Here the geometry becomes operational. Impose a location constraint or reference.

A concrete situation

Separate intercept from unit effects.

What the mathematics does not license

Unidentified shifts can look like multiple solutions. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Check rank. 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. Unbalanced panels

This step deserves a careful pause. Allow different visit counts.

A concrete situation

Weighting may change influence.

What the mathematics does not license

Large clusters can dominate. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Report distribution. 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. Missingness

The useful question is not only how to calculate, but what the calculation assumes. Describe dropout and intermittent gaps.

A concrete situation

Use sensitivity analysis.

What the mathematics does not license

Missing at random is an assumption. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Do not silently delete. 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. Uncertainty

A small example makes the mechanism visible. Cluster by panel unit or use justified dependence methods.

A concrete situation

Resample whole units when appropriate.

What the mathematics does not license

IID row bootstrap breaks dependence. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Mirror 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.

16. Claim

This is where modelling choices begin to affect evidence. State tau, effects, penalty and panel support.

A concrete situation

Teaching losses are zero and three.

What the mathematics does not license

Do not call repeated association a personal destiny. A trustworthy workflow records the choice before seeing the final test result.

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 regression for longitudinal data studies repeated observations on the same person, school, machine or organisation while allowing covariate effects to differ across conditional quantiles. Individual effects can absorb stable level differences. Koenker proposed regularising the many individual effects so that the common slope is not overwhelmed by one parameter for every unit. The method connects check loss, fixed effects, shrinkage and dependence in one transparent objective.

Consider two students observed twice. Student A has outcomes 2 and 3; student B has 5 and 6; time x is 0 then 1. At tau=.5, individual intercepts alpha_A=2 and alpha_B=5 with common slope beta=1 give zero residuals. A pooled model with intercept 3.5 and slope 1 gives residuals -1.5,-1.5,1.5,1.5 and median check-loss total 3. The individual effects recover the different baselines, but with only two observations per student the perfect fit also warns how easily fixed effects can absorb noise.

Penalising the absolute individual effects, or their deviations under a chosen parameterisation, shrinks them toward a common level and trades bias for variance. The penalty depends on scale and identification constraints. Repeated observations are not independent rows: serial correlation, cluster size, missing visits and informative dropout affect uncertainty and interpretation. A larger number of people does not make each person richly observed.

Report the panel unit, time index, tau, common and individual effects, normalisation, penalty, tuning rule, within-unit dependence, unbalanced panels, missingness and cluster-aware uncertainty. A quantile slope is conditional on the fitted model and population; it does not mean every unit changes by that amount or that the same named individual remains at one rank over time.

Frequently asked questions

Why use individual effects?

They absorb stable unit-specific levels that pooled models can mix into common slopes.

Why penalise them?

Shrinkage can reduce the variability created by estimating one effect per unit.

Are repeated rows independent?

No. Measurements from one unit usually share dependence.

What does an unbalanced panel change?

Units contribute different numbers of observations and may have different influence.

How should uncertainty be estimated?

With cluster-aware or model-appropriate methods that preserve the panel design.

What must be reported?

Panel unit, tau, effects, normalisation, penalty, missingness, dependence and validation.

Primary source and fact-check note

Roger Koenker, “Quantile Regression for Longitudinal Data,” Journal of Multivariate Analysis 91(1), 74–89 (2004), DOI 10.1016/j.jmva.2004.05.006. 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 observational unit, treatment or predictors, panel IDs, time order, response orientation, tail direction, missingness and dependence.

Objective audit

Write structural moments, check losses, individual-effect penalties, tail normalisations or multi-parameter likelihoods explicitly.

Assumption audit

List exclusion, relevance, rank, panel, tail-equivalence, response-family and link assumptions that the calculation needs.

Tuning audit

Record tau, penalty, threshold, tail sequence, family, smoothness and solver settings before interpreting output.

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.

Computation audit

Check moment balance, identification rank, optimisation ties, convergence, effective degrees of freedom and numerical sensitivity.

Uncertainty audit

Use weak-identification, cluster, extreme-value, simulation or likelihood methods matched to the actual design.

Validation audit

Use untouched units or future observations for loss, centile coverage, tail calibration and subgroup checks.

Final claim

Translate the conditional result without turning an assumption, percentile, fitted distribution or association into a 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: Panel unit

Begin with this mechanism: Name the repeated entity. Recreate the concrete case: Use students A and B. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Rows within a unit are dependent. Complete the transfer step: Keep IDs. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Time

Begin with this mechanism: Order visits 0 and 1. Recreate the concrete case: Observe each student twice. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Two visits provide little within-unit information. Complete the transfer step: Record timing. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Outcomes

Begin with this mechanism: Use A:2,3 and B:5,6. Recreate the concrete case: Student B has a higher stable level. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Values are illustrative only. Complete the transfer step: Keep units. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Target

Begin with this mechanism: Choose tau=.5. Recreate the concrete case: Fit a conditional median. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Another tau can give another slope. Complete the transfer step: Use grid later. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. 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: Individual effects

Begin with this mechanism: Set alpha_A=2 and alpha_B=5. Recreate the concrete case: Absorb baseline difference. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One effect per unit increases dimension. Complete the transfer step: Normalise. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Common slope

Begin with this mechanism: Set beta=1 for time. Recreate the concrete case: Predict both second outcomes. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A time trend need not be causal. Complete the transfer step: Keep interpretation. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Residuals

Begin with this mechanism: Subtract fitted values. Recreate the concrete case: All four residuals are zero. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Perfect fit can signal saturation. Complete the transfer step: Check degrees of freedom. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Pooled model

Begin with this mechanism: Use intercept 3.5 and slope 1. Recreate the concrete case: Ignore unit levels. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Pooling mixes within and between variation. Complete the transfer step: Compare explicitly. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. 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: Pooled residuals

Begin with this mechanism: Get -1.5,-1.5,1.5,1.5 in unit-time order. Recreate the concrete case: Absolute residual sum is 6. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Order must remain auditable. Complete the transfer step: Keep table. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Pooled loss

Begin with this mechanism: Multiply by .5 at the median. Recreate the concrete case: Total check loss is 3. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Training loss alone cannot choose a model. Complete the transfer step: Validate. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Penalty

Begin with this mechanism: Shrink individual effects under a stated parameterisation. Recreate the concrete case: Tune lambda. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Penalty changes the estimand and bias. Complete the transfer step: Publish formula. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Identification

Begin with this mechanism: Impose a location constraint or reference. Recreate the concrete case: Separate intercept from unit effects. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Unidentified shifts can look like multiple solutions. Complete the transfer step: Check rank. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. 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: Unbalanced panels

Begin with this mechanism: Allow different visit counts. Recreate the concrete case: Weighting may change influence. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Large clusters can dominate. Complete the transfer step: Report distribution. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Missingness

Begin with this mechanism: Describe dropout and intermittent gaps. Recreate the concrete case: Use sensitivity analysis. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Missing at random is an assumption. Complete the transfer step: Do not silently delete. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Uncertainty

Begin with this mechanism: Cluster by panel unit or use justified dependence methods. Recreate the concrete case: Resample whole units when appropriate. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: IID row bootstrap breaks dependence. Complete the transfer step: Mirror design. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. 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 tau, effects, penalty and panel support. Recreate the concrete case: Teaching losses are zero and three. Write panel IDs, times, outcomes, predictors, quantile levels, individual effects, common slopes, residuals, check losses, penalties, visit counts and original units with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call repeated association a personal destiny. Complete the transfer step: Bound conclusion. Change exactly one observation, treatment assignment, instrument, rank restriction, panel visit, individual effect, penalty, tau, sample size, threshold, tail index, family, link, scale, shape parameter, smoother, solver tolerance, bootstrap rule or validation sample. Explain which mathematical object changes, which estimator is selected, which conclusion survives and which claim must weaken. Keep design, objective, assumptions, tuning, computation, 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 weak identification, informative dropout, sparse tails, family misspecification, subgroup miscalibration, measurement error or high leverage. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.