Choose your reading route
eduKateSG · Why Mathematics? · Bayesian Bootstrap: Dirichlet Weights and Posterior Functionals
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 uncertainty is represented by random probability weights rather than repeated integer counts? The Bayesian bootstrap assigns a Dirichlet-distributed weight vector to the observed values. Each draw creates a complete probability distribution supported on the data, and the statistic becomes a posterior draw of a population functional.
Did You Know? Independent exponential numbers divided by their sum produce Dirichlet(1,…,1) weights. This turns a simple random-number calculation into a distribution over all positive weight vectors that sum to one.
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
The ordinary bootstrap samples rows with replacement and produces multinomial counts. The Bayesian bootstrap keeps every distinct observed support point and randomises its probability mass continuously.
Rubin introduced the Bayesian bootstrap as a Bayesian analogue of the bootstrap that simulates a posterior distribution for a parameter functional. He also stressed sensitivity to modelling assumptions. Read the published paper copy.
For n observations, draw W=(W1,…,Wn) from Dirichlet(1,…,1). Every Wi is nonnegative and the weights sum exactly to one. A weighted statistic is evaluated for each draw.
Suppose four exponential draws are (0.2,0.5,0.1,0.2). Their sum is 1.0, so the resulting Dirichlet weights are the same four values: (0.2,0.5,0.1,0.2).
| Exponentials | Sum | Weights |
|---|---|---|
| 0.2,0.5,0.1,0.2 | 1.0 | 0.2,0.5,0.1,0.2 |
For values (2,4,6,8), those weights give 0.2×2+0.5×4+0.1×6+0.2×8=4.6. A new Dirichlet draw gives another posterior draw of the mean functional.
| Values | Weights | Mean |
|---|---|---|
| 2,4,6,8 | 0.2,0.5,0.1,0.2 | 4.6 |
Independent Gamma variables normalised by their total have a Dirichlet distribution. Exponential(1) is Gamma(shape 1, rate 1), producing the uniform Dirichlet used in Rubin’s basic construction.
The distribution of weighted functionals is a posterior under a particular nonparametric prior construction. It should not be described as assumption-free or automatically equivalent to a fully specified Bayesian model.
Ordinary multinomial counts are discrete and some observations are omitted. Dirichlet weights are continuous and positive with probability one, so every observed support point normally remains represented.
Weighted quantiles require a clear convention for cumulative weights and ties. The Bayesian bootstrap median can jump between observed values even though the weights vary continuously.
Weighted estimating equations can produce posterior draws of regression functionals. Software must interpret weights as probability or frequency weights consistently with the estimator.
The basic method places mass only at observed points. It cannot generate unobserved values or extrapolate tail shape without an additional model.
Dirichlet parameters other than one change concentration and prior influence. Calling every random-weight analysis a Bayesian bootstrap hides a material modelling choice.
For Dirichlet(1,…,1), E(Wi)=1/n. The variance is (n−1)/(n²(n+1)), and different weights are negatively correlated because they must sum to one.
With n=4, E(Wi)=0.25 and Var(Wi)=3/(16×5)=0.0375. The standard deviation of one weight is about 0.1936, showing that individual masses can vary substantially.
Posterior quantiles of the weighted statistic form Bayesian-bootstrap credible intervals. Their repeated-sampling coverage is a separate property that should be studied rather than assumed.
Random weighted means can propagate uncertainty into a cost-benefit decision. The decision still requires utilities, constraints and domain evidence beyond the bootstrap weights.
Survey weights encode design or adjustment information. Dirichlet posterior weights represent uncertainty over the empirical support. They are not interchangeable, though specialised methods can combine them.
Both use random weights, but the Bayesian bootstrap uses positive weights summing to one and has a posterior interpretation. Multiplier processes often use mean-zero weights and target sampling distributions.
The methods can give similar first-order uncertainty for smooth statistics, yet differ in discreteness, omission patterns and interpretation. Report which law generated the weights.
Check weight sums and marginal behaviour, reproduce simple functionals, compare with ordinary bootstrap, simulate frequentist coverage when relevant, and inspect sensitivity to prior concentration and outliers.
The method inherits empirical support, independence assumptions and functional instability. It does not repair biased data, dependence, missing populations or unidentified causal effects.
Place one weight draw per row, verify row sums, calculate weighted statistics and chart the posterior distribution. Keep any survey or precision weights in separate labelled columns.
Generate exponentials, normalise by their sum and unit-test weights (0.2,0.5,0.1,0.2), mean 4.6 and n=4 variance 0.0375. Save seeds and weight semantics.
Bayesian-bootstrap ideas appear in econometrics, risk, medicine and probabilistic machine learning. Careers require Bayesian reasoning, numerical computing and careful interpretation.
Compare ordinary and Bayesian bootstrap uncertainty for a mean, median and weighted regression. Add one outlier and study how continuous weights change sensitivity.
Bayesian bootstrap does not mean resampling from a fitted parametric posterior, random weights are not survey weights, and a credible interval is not automatically a frequentist confidence interval.
CHAPTER 27 OF 34 · Explore applications
27. Connection to recent mathematics articles
The importance-resampling guide explains correction weights. The multiplier-bootstrap guide explains mean-zero random weighting.
Students should normalise four exponentials and calculate one weighted mean. Parents can ask what probability model gives the weights and what interpretation the interval has.
- Normalise the four draws.
- Recompute 4.6.
- Calculate E(Wi).
- Calculate Var(Wi) for n=4.
- Compare with multinomial counts.
### Do weights sum to one? Yes.
### Can an observed point receive exactly zero weight? With continuous Dirichlet weights, probability zero.
### Is this a posterior method? Yes, under its construction.
### Is it assumption-free? No.
Report observations, functional, Dirichlet parameters, B, seed, weight convention, weighted-estimator semantics, interval interpretation, sensitivity checks and support limitations.
The weight vector lives on the n−1 dimensional simplex. The sum constraint creates negative dependence between coordinates and ensures each draw is a complete discrete probability distribution.
Confirm positive unit-sum weights, stated Dirichlet parameters, correct weighted functional, honest posterior interpretation, sensitivity to support and no confusion with design weights.
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 observations
Setup: Start from the article’s Bayesian-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: Draw exponentials
Setup: Start from the article’s Bayesian-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: Sum exponentials
Setup: Start from the article’s Bayesian-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: Normalise weights
Setup: Start from the article’s Bayesian-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: Check simplex
Setup: Start from the article’s Bayesian-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 weighted mean
Setup: Start from the article’s Bayesian-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: Draw new weights
Setup: Start from the article’s Bayesian-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: Calculate posterior quantile
Setup: Start from the article’s Bayesian-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: Change Dirichlet concentration
Setup: Start from the article’s Bayesian-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: Calculate expected weight
Setup: Start from the article’s Bayesian-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: Calculate weight variance
Setup: Start from the article’s Bayesian-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 median
Setup: Start from the article’s Bayesian-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 regression
Setup: Start from the article’s Bayesian-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: Add outlier
Setup: Start from the article’s Bayesian-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: Compare ordinary bootstrap
Setup: Start from the article’s Bayesian-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: Compare multiplier weights
Setup: Start from the article’s Bayesian-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: Separate survey weights
Setup: Start from the article’s Bayesian-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: Check support
Setup: Start from the article’s Bayesian-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: Increase B
Setup: Start from the article’s Bayesian-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: Change seed
Setup: Start from the article’s Bayesian-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: Plot posterior
Setup: Start from the article’s Bayesian-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: Test frequentist coverage
Setup: Start from the article’s Bayesian-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 interpretation
Setup: Start from the article’s Bayesian-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 Bayesian-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 Bayesian-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.