Statistical Methods

Diagnose a Degenerate BCa Bootstrap Interval

Define the task

A bias-corrected and accelerated bootstrap interval uses information from the resampled statistic and jackknife changes. If the statistic hardly varies, the required corrections can become undefined or unstable. Before interpreting an interval, record the estimand, resampling unit, confidence level, method, number of resamples and random generator. Repeated observations within a participant cannot be treated as independent rows merely because a bootstrap function accepts an array.

Carry out the check

Inspect the distribution of resampled statistics and the leave-one-out estimates. Current documentation warns that BCa limits can be NaN for degenerate distributions. More repetitions cannot create information absent from the observed data. A different interval method may be appropriate, but choose it for statistical reasons and disclose the change; do not replace failed limits with zeros. If data are constant because of rounding, censoring or a programming mistake, investigate that mechanism first. Distinguish numerical failure from evidence of zero population uncertainty.

Worked hypothetical example

For hypothetical values (5,5,5,5), every ordinary resample has mean 5 and every leave-one-out mean is also 5. The percentile interval collapses to (5,5), while the acceleration calculation has zero variation in its denominator. Neither outcome proves the population mean is known without uncertainty. Keep this synthetic case as an edge test. A bootstrap report should include diagnostics and unresolved limitations rather than only two apparently precise endpoints.

Method checked 18 November 2026 against Primary method documentation: bootstrap.

Related reading

Reading path: Analyze data. Guide 171 of 172. Choose any guide to suit your project.

Statistical Methods

Read a Bootstrap Percentile Interval as a Method Choice

Distinguish percentile limits from standard-error intervals and inspect instability before reporting them.

Statistical Methods

Compute an interval probability under a bounded uniform model

Calculate interval overlap under a bounded continuous distribution. Includes a worked hypothetical example and checks for parameter-transfer errors.

Browse this reading pathTools for this stageGuides and downloads for this stage