Choose your reading route
eduKateSG · Why Mathematics? · Sieve Bootstrap: Autoregressive Approximations for Time Series
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 a time series has dependence extending beyond any one chosen block length? The autoregressive sieve bootstrap fits an AR(p) approximation, resamples centred residuals, and recursively generates new series. As the order p grows slowly with sample size, the fitted sieve can approximate a broad class of linear processes.
Did You Know? The method is called a sieve because the approximating model becomes richer with more data: p may increase toward infinity while remaining much smaller than n.
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
Independent resampling destroys serial dependence. Fixed blocks preserve only observed local segments. A sieve instead estimates a dynamic filter and generates fresh paths through that filter.
Bühlmann studied an autoregressive sieve with order p increasing more slowly than n and proved consistency for classes of estimators. Read the Berkeley technical report record.
For a centred series, estimate X_t=a1 X_{t−1}+…+ap X_{t−p}+e_t. Order selection, intercept or trend treatment and stationarity checks are part of the resampling method.
Let a=0.6, previous value X-star=5 and resampled residual e-star=−1. The next bootstrap value is 0.6×5−1=2.
| a | Previous | Innovation | Next |
|---|---|---|---|
| 0.6 | 5 | -1 | 2 |
If the next residual is 0.5, then the following value is 0.6×2+0.5=1.7. Each generated value becomes input for the next step.
| a | Previous | Innovation | Next |
|---|---|---|---|
| 0.6 | 2 | 0.5 | 1.7 |
Estimated residuals are normally centred before resampling so generated innovations have mean zero. Leaving a residual mean can create artificial drift.
Starting from observed values can carry initial-condition effects. Simulating a burn-in and discarding it helps the generated path approach the fitted stationary distribution.
AIC, BIC or a theoretical rule may select p. Too small an order misses dependence; too large an order creates noisy coefficients and unstable recursion.
Asymptotic theory lets p=p(n) grow, expanding the approximation while controlling estimation error. Keeping p fixed forever and calling it a sieve can misstate the method.
Estimated AR roots should imply a stable recursion for stationary targets. Near-unit roots or structural breaks may need specialised bootstraps rather than forced stabilisation.
Deterministic trend and seasonal effects should be modelled or removed consistently, then restored. Residual resampling alone does not preserve calendar structure.
Simple residual resampling assumes innovations are identically distributed and sufficiently independent. Conditional heteroscedasticity calls for a wild, model-based or volatility-aware innovation scheme.
Generate many paths of length n, calculate each sample mean and use their spread for uncertainty. The fitted AR filter carries estimated autocorrelation into the replicates.
AR approximations imply a spectral density. A well-chosen sieve can reproduce broad frequency behaviour, while a poor order can invent peaks or erase cycles.
Bootstrap paths can propagate parameter and innovation uncertainty into forecasts. Refitting the AR model inside each path targets more uncertainty than simulating only innovations from one fixed fit.
Moving-block bootstrap resamples contiguous observed segments. Sieve bootstrap generates new paths from an estimated global AR approximation; one relies on blocks, the other on a growing model.
A fixed low-order AR parametric bootstrap assumes that finite model is correct. The sieve interpretation treats increasing AR order as an approximation to a broader process class.
Stationary bootstrap concatenates random-length observed blocks. Sieve paths are recursive and need not contain any original segment verbatim.
Compare autocorrelation, partial autocorrelation, spectrum, residual distribution and path stability between observed and bootstrap series. Matching one lag is insufficient.
Simulate processes with known dependence, vary p and selection rule, check coverage, include burn-in sensitivity, record unstable fits and compare block alternatives.
Sieve bootstrap can fail with structural breaks, nonlinear dynamics, long memory outside the approximation rate, nonstationarity, volatility clustering or poor residual exchangeability.
List time, observed value, fitted value, centred residual, resampled residual and generated value. Add sheets comparing ACF and spectra across p choices.
Fit AR(p), centre residuals, simulate burn-in and unit-test 0.6×5−1=2 and 0.6×2+0.5=1.7. Save order rule, seed and stability diagnostics.
Sieve methods matter in forecasting, economics, climate, engineering and signal processing. Careers require time-series theory, numerical stability and domain knowledge.
Simulate ARMA data, compare sieve and moving-block intervals for the mean over several n and p rules, then add a structural break to reveal failure.
A high p does not guarantee realism, residual resampling is not valid for every volatility pattern, and a recursive path is not an observed history.
CHAPTER 27 OF 34 · Explore applications
27. Connection to recent mathematics articles
The moving-block-bootstrap guide explains observed blocks. The stationary-bootstrap guide explains random-length blocks.
Students should generate two recursive values by hand. Parents can ask how p was chosen and whether the generated series reproduces observed dependence without becoming unstable.
- Calculate both recursion steps.
- Explain residual centring.
- Diagnose an unstable AR root.
- Compare burn-in choices.
- Compare sieve and blocks.
### What is fitted? An autoregressive approximation.
### What is resampled? Usually centred residuals.
### Does p stay fixed in sieve theory? It can grow with n.
### Are paths copied blocks? No.
Report n, preprocessing, p rule, coefficient method, residual handling, burn-in, B, seed, stability failures, dependence diagnostics, target statistic and block benchmark.
Each innovation affects future values through AR coefficients. The impulse response explains how residual resampling becomes serially correlated path uncertainty.
Confirm stationary target, justified p, centred exchangeable innovations, stable recursion, adequate burn-in, dependence diagnostics and honest comparison with block methods.
Continue through the Mathematics Learning Hub. Reproduce one cluster, stratum, autoregressive or spatial-block calculation by hand before trusting a large automated run.
Extended practice laboratory
Practice case 1: Centre series
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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 p
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Fit AR coefficients
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Calculate residuals
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Centre residuals
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Draw innovation
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Generate first value
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Generate second value
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Run burn-in
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Generate n values
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Calculate mean
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Check ACF
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Check spectrum
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Increase p
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Reduce p
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Use AIC
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Use BIC
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Add seasonality
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Add trend
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Add volatility
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Compare moving blocks
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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: Compare stationary blocks
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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 stability
Setup: Start from the article’s sieve-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 grouping, recursion or geometry 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 sieve-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 grouping, recursion or geometry 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 sieve-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 grouping, recursion or geometry 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.