NEWS.md

In funbootband: Simultaneous Prediction and Confidence Bands for Functional Data

funbootband 0.3.0

• Replaced the former curve-by-curve hierarchical sampler with an intact-subject cluster bootstrap. Subjects are sampled with replacement and all observed curves belonging to a selected subject are retained.
• Defined the clustered prediction target as one future curve from a new subject. Subjects are weighted equally and curves are weighted equally within subject, including for unequal cluster sizes.
• Prediction calibration now retains one supremum statistic per pseudo-future curve instead of maximizing over the whole observed collection, and uses a replicate-specific pointwise scale.
• Clustered confidence bands now use subject mean curves for their pointwise standard errors, resample intact subjects, and use replicate-specific studentization.
• The returned metadata now records the estimand, weighting convention, bootstrap unit, cluster sizes, and Fourier-reconstructed curve representation.
• Documentation and examples now distinguish the revised intact-subject method from the hierarchical procedure described in Koska et al. (2023).
• Added numerical reference tests for the Rcpp kernels, unequal-cluster examples, edge-case tests, and a reproducible coverage-validation script.

funbootband 0.2.0

funbootband 0.1.1 (2025-09-22)

• Add Fourier preprocessing in band() via k.coef (default 50) to honor smooth/periodic structure.
• New vignette: “funbootband: Simultaneous Bands for Functional Data”.
• Faster, cleaned bootstrap pipeline (Rcpp); small B used in examples/tests for CRAN timing.
• Documentation and tests refreshed.

funbootband 0.1.0 (2025-08-01)

• Initial release: simultaneous prediction and confidence bands with clustered bootstrap.

funbootband documentation built on Sept. 18, 2026, 5:08 p.m.