Choose your reading route
eduKateSG · Why Mathematics? · Smoothed Bootstrap: Kernel Noise, Bandwidths and Continuous Support
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 8–14
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 15–21
Check assumptions
See where design and calibration matter.Route 4 · Chapters 22–28
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 29–34
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important when an empirical distribution is too discrete to represent a genuinely continuous population? The smoothed bootstrap samples an observed value and then adds calibrated kernel noise. The bandwidth controls how much probability mass spreads between observations, trading discreteness against smoothing bias.
Did You Know? If X-star=X_i+hZ with independent kernel noise Z of mean zero and variance one, then Var(X-star) equals the empirical variance under the sampling convention plus h². Smoothing changes spread unless a variance correction is made.
Full chapter index
Chapters 1–7 · Understand the question
Chapters 8–14 · Reproduce the example
Chapters 15–21 · Check assumptions
Chapters 22–28 · Explore applications
Chapters 29–34 · Study and verify
Ordinary nonparametric bootstrap samples only observed values. For density, quantile and other nonsmooth problems, that discrete support can create artificial ties or poor approximation.
Silverman and Young developed criteria for deciding whether smoothing improves bootstrap estimation rather than assuming it always helps. Read the Biometrika abstract.
Draw an index I uniformly from the observations, draw kernel noise Z, and set X-star=X_I+hZ. The bandwidth h has the same units as X.
If the selected observation is 10, h=2 and the kernel draw is Z=−0.5, then X-star=10+2(−0.5)=9.
| Observed | Bandwidth | Noise | Smoothed value |
|---|---|---|---|
| 10 | 2 | -0.5 | 9 |
If the empirical component has variance 9 and h=2 with Var(Z)=1, the smoothed draw has variance 9+4=13 under independence. Ignoring this inflation can distort a scale estimate.
| Empirical component | Added kernel variance | Total |
|---|---|---|
| 9 | 4 | 13 |
Gaussian noise is common, but bounded kernels can respect local support better. Kernel moments and boundary behaviour matter more than decorative curve shape.
A small h resembles the ordinary bootstrap; a large h oversmooths modes, boundaries and tails. Data-driven density bandwidths are not automatically optimal for every target statistic.
Smoothing can reduce artifacts from a discrete empirical distribution but introduces bias by altering the estimated population. The best h balances these errors for the target functional.
Some smoothed-bootstrap schemes rescale the jittered data so the variance matches a desired estimate. The correction must be explicit because it changes both centre and spread calculations.
Adding symmetric noise to nonnegative data can create impossible negative values. Reflection, transformation, boundary kernels or a scientifically motivated parametric tail may be needed.
Angles require circular kernels so values near 0 and 360 degrees remain neighbours. Adding ordinary Gaussian noise on the number line breaks the geometry.
A bandwidth matrix controls scale and correlation of multivariate noise. Independent coordinate jitter can erase or invent geometric structure.
Smoothing reduces repeated bootstrap quantiles caused by ties and can improve approximation for some distributions. It may also shift tail quantiles if h is too large.
For a kernel density estimator, a smoothed bootstrap can reproduce continuous sampling variation more naturally than sampling only observed points. Bandwidth uncertainty remains part of the problem.
If recorded values are rounded to the nearest unit, carefully chosen smoothing can represent latent continuity. The bandwidth should reflect measurement resolution and the inferential target, not simply make a nicer graph.
Failure times are nonnegative and may be censored. Naive additive smoothing can violate both support and censoring, requiring specialised survival resampling.
Ordinary bootstrap has no smoothing bandwidth and preserves empirical support exactly. Smoothed bootstrap adds a model of local continuity and therefore another assumption.
A parametric bootstrap simulates from a fitted global probability family. A smoothed bootstrap uses a local kernel mixture centred at observed values.
Bayesian bootstrap randomises masses on fixed observed support. Smoothed bootstrap changes the support by generating new nearby values.
Vary h over a prespecified grid, inspect support violations, compare bias and coverage in simulations, test variance correction, and benchmark against ordinary and parametric alternatives.
Smoothing can invent impossible values, erase multimodality, distort tails and correlations, or hide rounding mechanisms. It cannot repair selection bias or wrong sampling units.
Store selected index, observed value, kernel draw, bandwidth, jittered value and any boundary or variance correction. Plot results across a bandwidth grid.
Unit-test 10+2(−0.5)=9 and variance 9+4=13. Use separate functions for kernel generation, support correction and statistic calculation; save h and seed.
Smoothed resampling appears in density estimation, quantiles, measurement science and reliability. Careers require probability, numerical methods and substantive support constraints.
Simulate rounded continuous data, compare ordinary and smoothed bootstrap quantile intervals over bandwidths, and measure ties, bias, coverage and invalid values.
Smoothing is not cosmetic graphing, h is not unitless, Gaussian noise is not universally appropriate, and fewer ties do not prove better inference.
CHAPTER 27 OF 34 · Explore applications
27. Connection to recent mathematics articles
The parametric-bootstrap guide explains global fitted simulation. The Bayesian-bootstrap guide keeps fixed support.
Students should calculate one jittered value and variance change. Parents can ask what h means in real units and how impossible generated values were prevented.
- Calculate the jittered value.
- Calculate added variance.
- Explain bandwidth units.
- Diagnose a boundary violation.
- Compare local and global models.
### Are new values generated? Yes.
### What controls smoothing? The bandwidth and kernel.
### Can smoothing change variance? Yes.
### Is more smoothing always better? No.
Report data support, kernel, bandwidth rule, boundary method, variance correction, B, seed, target statistic, sensitivity grid, support violations and simulation performance.
The smoothed empirical distribution is a convolution of the discrete empirical distribution with a scaled kernel. Convolution spreads each point mass into a local bump and adds kernel variance.
Confirm scientifically valid support, bandwidth units, kernel moments, variance effects, boundary handling, sensitivity across h and evidence that smoothing improves the target inference.
Continue through the Mathematics Learning Hub. Reproduce one weight, count or smoothing calculation by hand, then compare it with a tested implementation.
Extended practice laboratory
Practice case 1: List observed values
Setup: Start from the article’s smoothed-bootstrap mechanism for case 1. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 2: Choose kernel
Setup: Start from the article’s smoothed-bootstrap mechanism for case 2. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 3: Choose bandwidth
Setup: Start from the article’s smoothed-bootstrap mechanism for case 3. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 4: Draw index
Setup: Start from the article’s smoothed-bootstrap mechanism for case 4. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 5: Draw kernel noise
Setup: Start from the article’s smoothed-bootstrap mechanism for case 5. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 6: Calculate jittered value
Setup: Start from the article’s smoothed-bootstrap mechanism for case 6. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 7: Calculate added variance
Setup: Start from the article’s smoothed-bootstrap mechanism for case 7. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 8: Repeat B times
Setup: Start from the article’s smoothed-bootstrap mechanism for case 8. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 9: Reduce bandwidth
Setup: Start from the article’s smoothed-bootstrap mechanism for case 9. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 10: Increase bandwidth
Setup: Start from the article’s smoothed-bootstrap mechanism for case 10. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 11: Check negative values
Setup: Start from the article’s smoothed-bootstrap mechanism for case 11. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 12: Use reflection
Setup: Start from the article’s smoothed-bootstrap mechanism for case 12. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 13: Use log transform
Setup: Start from the article’s smoothed-bootstrap mechanism for case 13. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 14: Use circular kernel
Setup: Start from the article’s smoothed-bootstrap mechanism for case 14. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 15: Use bandwidth matrix
Setup: Start from the article’s smoothed-bootstrap mechanism for case 15. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 16: Estimate quantile
Setup: Start from the article’s smoothed-bootstrap mechanism for case 16. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 17: Estimate density
Setup: Start from the article’s smoothed-bootstrap mechanism for case 17. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 18: Model rounding
Setup: Start from the article’s smoothed-bootstrap mechanism for case 18. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 19: Compare ordinary bootstrap
Setup: Start from the article’s smoothed-bootstrap mechanism for case 19. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 20: Compare parametric bootstrap
Setup: Start from the article’s smoothed-bootstrap mechanism for case 20. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 21: Change seed
Setup: Start from the article’s smoothed-bootstrap mechanism for case 21. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 22: Validate coverage
Setup: Start from the article’s smoothed-bootstrap mechanism for case 22. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 23: Audit support
Setup: Start from the article’s smoothed-bootstrap mechanism for case 23. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 24: Name limits
Setup: Start from the article’s smoothed-bootstrap mechanism for case 24. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.
Practice case 25: Write conclusion
Setup: Start from the article’s smoothed-bootstrap mechanism for case 25. Task: Change the named quantity, recompute the dependent result and compare with the original.
Check: Preserve the sampling unit, verify the weight, count or smoothing identity, and state one diagnostic before interpretation. Show the arithmetic, name the response and predictor units, and separate the numerical result from the claim it supports. Then vary one defensible assumption, calculate again and explain whether the conclusion changed because of the data, the fitted model or the validation purpose. Finish with one compact table or graph that could reveal a failure hidden by the summary.