WHY MATHEMATICS?

Why Mathematics? | Bayesian Spatial Quantile Regression: Distance Decay, Spatial Priors, Conditional Percentile Maps and Site-Wise Calibration

How nearby monitoring sites can share information while lower, middle and upper conditional outcomes vary across a region, and why map smoothness needs 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

Record coordinates, monitoring design, dates and covariates
Model conditional quantiles with spatially varying distributional structure
Borrow information through an explicit spatial prior or basis
Validate at held-out sites and show uncertainty with every map

Chapter map

1. Locations
2. Quantile target
3. Distance
4. Correlation model
5. Near pair
6. Middle pair
7. Far pair
8. Spatial prior
9. Covariates
10. Knots or basis
11. Monitoring design
12. Interpolation
13. Extrapolation
14. Validation
15. Computation
16. Claim

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

1. Locations

This is where modelling choices begin to affect evidence. Record coordinates and projection.

A concrete situation

Use 0,1,3 km on a line.

What the mathematics does not license

Latitude degrees are not kilometres. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

State CRS. 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. Quantile target

The central mathematical idea is concrete. Choose conditional tau levels.

A concrete situation

Map median and upper tail.

What the mathematics does not license

Different tails have different support. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Report grid. 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. Distance

Here the geometry becomes operational. Compute pairwise separations.

A concrete situation

Distances are 1,2,3 km.

What the mathematics does not license

Barriers can defeat Euclidean distance. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Compare metric. 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. Correlation model

This step deserves a careful pause. Use R(d)=exp(-d/r) for illustration.

A concrete situation

Set r=2 km.

What the mathematics does not license

This kernel is a modelling choice. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Test alternatives. 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. Near pair

The useful question is not only how to calculate, but what the calculation assumes. Evaluate exp(-1/2).

A concrete situation

Get .6065.

What the mathematics does not license

Correlation is not equality. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

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

6. Middle pair

A small example makes the mechanism visible. Evaluate exp(-2/2).

A concrete situation

Get .3679.

What the mathematics does not license

Location labels must match matrix entries. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Audit 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.

7. Far pair

This is where modelling choices begin to affect evidence. Evaluate exp(-3/2).

A concrete situation

Get .2231.

What the mathematics does not license

Long-range structure may remain. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Inspect variogram. 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. Spatial prior

The central mathematical idea is concrete. Smooth site-specific distributional quantities.

A concrete situation

Borrow more between close sites.

What the mathematics does not license

Prior smoothness can mask discontinuities. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Map uncertainty. 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. Covariates

Here the geometry becomes operational. Include weather or land-use variables measured before prediction.

A concrete situation

Keep units and timestamps.

What the mathematics does not license

Spatial confounding remains possible. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Use design knowledge. 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. Knots or basis

This step deserves a careful pause. Approximate a large spatial field.

A concrete situation

Place basis support over domain.

What the mathematics does not license

Poor placement misses local features. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Run refinement. 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. Monitoring design

The useful question is not only how to calculate, but what the calculation assumes. Audit where sensors exist.

A concrete situation

Compare urban and remote coverage.

What the mathematics does not license

Preferential sampling biases maps. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Model or qualify. 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. Interpolation

A small example makes the mechanism visible. Predict inside observed domain.

A concrete situation

Use nearby evidence.

What the mathematics does not license

Smooth prediction is not direct measurement. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Label estimate. 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. Extrapolation

This is where modelling choices begin to affect evidence. Identify sites beyond support.

A concrete situation

Widen uncertainty.

What the mathematics does not license

A prior cannot create missing regimes. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Bound map. 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. Validation

The central mathematical idea is concrete. Hold out whole locations.

A concrete situation

Assess quantile coverage and check loss.

What the mathematics does not license

Random rows leak spatial neighbours. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Use spatial folds. 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. Computation

Here the geometry becomes operational. Check chains or approximation error.

A concrete situation

Inspect convergence and sensitivity.

What the mathematics does not license

Large datasets can force approximations. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Report method. 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 step deserves a careful pause. State dates, region, prior and support.

A concrete situation

Teaching correlations are .6065,.3679,.2231.

What the mathematics does not license

Do not infer policy effects from a map alone. That distinction matters because a correct formula can still answer the wrong operational question.

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

Bayesian spatial quantile regression models how different conditional percentiles vary across a map. Reich, Fuentes and Dunson developed a spatial model for an entire conditional distribution, allowing covariate effects and residual distributional shape to vary from site to site and smoothing related quantities through spatial priors. Their ozone application illustrates why a mean map can miss changes in upper-tail pollution that matter for planning and health research.

For an illustrative exponential spatial correlation R(d)=exp(-d/r), take sites at 0,1 and 3 kilometres with range r=2 kilometres. The correlations are exp(-.5)=.6065 between the first two sites, exp(-1)=.3679 between the last two, and exp(-1.5)=.2231 between the first and third. Nearby sites therefore borrow more information in this teaching kernel. These numbers do not reproduce every component of the paper; they show the distance-decay idea behind one common spatial prior.

Spatial smoothness is an assumption, not a fact supplied by coordinates. Barriers, coastlines, prevailing winds, road networks and measurement practices can make two nearby sites behave differently. Preferential monitoring locations can bias the observed spatial distribution. A posterior map may appear smooth even where data are sparse, so uncertainty, cross-validation by location and sensitivity to the covariance model must travel with the map.

Report coordinate system, distance units, monitoring design, response definition, covariates, quantile levels, basis or knot structure, spatial covariance or prior, hyperpriors, computation, convergence and held-out-site calibration. Separate interpolation from extrapolation and observational association from causal climate or policy effects. Date environmental data and avoid extending one historical application beyond the period and region it measured.

Frequently asked questions

Why map quantiles instead of only means?

Upper and lower conditional outcomes can change differently across space.

What does the spatial prior do?

It formalises how related locations may borrow information.

Does proximity guarantee similarity?

No. Barriers, networks and local processes can break distance decay.

Why hold out whole sites?

Random-row validation leaks information from nearby observations.

What is preferential sampling?

Monitoring locations may be chosen in ways related to the process being mapped.

What must be reported?

Coordinates, dates, design, quantiles, prior, basis, computation, spatial validation and support.

Primary source and fact-check note

Brian J. Reich, Montserrat Fuentes and David B. Dunson, “Bayesian Spatial Quantile Regression,” Journal of the American Statistical Association 106(493), 6–20 (2011), DOI 10.1198/jasa.2010.ap09237. 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 outcomes, reference ranks, covariates, coordinates, monitoring locations, timestamps, sampling intervals, outcome and predictor quantile grids before estimation.

Objective audit

Write the transport assignment, convex-gradient constraint, kernel-weighted check loss, spatial conditional-distribution model or trigonometric quantile objective explicitly.

Tuning audit

Record rank grid, basis, kernel, bandwidth, spatial range, knots, priors, frequency grid, quantile grid, normalisation and stopping tolerance.

Computation audit

Check primal-dual gaps, cyclic monotonicity, effective local sample sizes, covariance conditioning, chain convergence, approximation error and frequency resolution.

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 sparse reference corners, predictor tails, unsampled regions, map boundaries, missing time blocks, short series and off-support predictions.

Uncertainty audit

Propagate rank discretisation, sample-quantile, bandwidth, spatial-prior, hyperparameter, dependence, frequency-search and solver uncertainty through the complete workflow.

Validation audit

Use untouched joint outcomes, blocked time windows or held-out locations for multivariate scores, check loss, quantile coverage, spatial calibration and future-cycle stability.

Final claim

Translate the conditional result without turning a vector rank, dependence surface, smooth percentile map or spectral peak into a causal or universal 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: Locations

Begin with this mechanism: Record coordinates and projection. Recreate the concrete case: Use 0,1,3 km on a line. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Latitude degrees are not kilometres. Complete the transfer step: State CRS. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Quantile target

Begin with this mechanism: Choose conditional tau levels. Recreate the concrete case: Map median and upper tail. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Different tails have different support. Complete the transfer step: Report grid. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Distance

Begin with this mechanism: Compute pairwise separations. Recreate the concrete case: Distances are 1,2,3 km. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Barriers can defeat Euclidean distance. Complete the transfer step: Compare metric. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Correlation model

Begin with this mechanism: Use R(d)=exp(-d/r) for illustration. Recreate the concrete case: Set r=2 km. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: This kernel is a modelling choice. Complete the transfer step: Test alternatives. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. 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: Near pair

Begin with this mechanism: Evaluate exp(-1/2). Recreate the concrete case: Get .6065. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Correlation is not equality. Complete the transfer step: Keep precision. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Middle pair

Begin with this mechanism: Evaluate exp(-2/2). Recreate the concrete case: Get .3679. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Location labels must match matrix entries. Complete the transfer step: Audit table. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Far pair

Begin with this mechanism: Evaluate exp(-3/2). Recreate the concrete case: Get .2231. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Long-range structure may remain. Complete the transfer step: Inspect variogram. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Spatial prior

Begin with this mechanism: Smooth site-specific distributional quantities. Recreate the concrete case: Borrow more between close sites. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Prior smoothness can mask discontinuities. Complete the transfer step: Map uncertainty. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. 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: Covariates

Begin with this mechanism: Include weather or land-use variables measured before prediction. Recreate the concrete case: Keep units and timestamps. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Spatial confounding remains possible. Complete the transfer step: Use design knowledge. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Knots or basis

Begin with this mechanism: Approximate a large spatial field. Recreate the concrete case: Place basis support over domain. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Poor placement misses local features. Complete the transfer step: Run refinement. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Monitoring design

Begin with this mechanism: Audit where sensors exist. Recreate the concrete case: Compare urban and remote coverage. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Preferential sampling biases maps. Complete the transfer step: Model or qualify. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Interpolation

Begin with this mechanism: Predict inside observed domain. Recreate the concrete case: Use nearby evidence. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Smooth prediction is not direct measurement. Complete the transfer step: Label estimate. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. 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: Extrapolation

Begin with this mechanism: Identify sites beyond support. Recreate the concrete case: Widen uncertainty. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A prior cannot create missing regimes. Complete the transfer step: Bound map. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Validation

Begin with this mechanism: Hold out whole locations. Recreate the concrete case: Assess quantile coverage and check loss. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Random rows leak spatial neighbours. Complete the transfer step: Use spatial folds. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Computation

Begin with this mechanism: Check chains or approximation error. Recreate the concrete case: Inspect convergence and sensitivity. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Large datasets can force approximations. Complete the transfer step: Report method. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. 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 dates, region, prior and support. Recreate the concrete case: Teaching correlations are .6065,.3679,.2231. Write coordinates, distance units, monitoring dates, responses, covariates, quantile levels, correlations, spatial ranges, basis weights, posterior summaries and coverage with units, signs, identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not infer policy effects from a map alone. Complete the transfer step: Bound conclusion. Change exactly one outcome coordinate, reference rank, transport assignment, covariate basis, theta level, tau level, sample quantile, kernel weight, bandwidth, monitoring site, spatial distance, correlation range, prior, sampling interval, frequency, phase, sine coefficient, cosine coefficient 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 unit-dependent transport, non-convex maps, sparse local tails, serial dependence, spatial barriers, preferential monitoring, irregular sampling, structural breaks or multiple frequency searches. State what extra measurement would separate the explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.