Resampling-Based Inference for Incomplete Paired Functional Data
Abstract
We consider testing equality of mean functions for paired functional data, where one entire function is missing. To use all available information from the paired and unpaired observations, we propose globalized test statistics that are based on a weighted sum of point-wise paired-t and Welch-t-type statistics. By the use of empirical process theory, we derive the asymptotic limit distributions of the resulting test statistics under the null hypothesis and show consistency against fixed alternatives. As they are non-pivotal and contain complex dependencies from the paired and incomplete observations, a permutation and parametric bootstrap procedures is proposed to compute critical values. The work includes a theoretical investigation of the asymptotic behaviour of the two resampling approaches and state regularity conditions under which the corresponding resampling tests are valid. Simulation studies show that the proposed methods provide accurate type I error control and competitive power for most settings under study. An illustrative analysis of functional emission data demonstrates the applicability of the methods.
Related
- Quadratic Functional Estimation and Pointwise Inference from Paired Measurements with Unknown Errors Under Polynomial Lower Bounds
- Bayesian Inference for Incomplete 2 × 2 Diagnostic Tables
- Inference for Directions in Canonical Correlation Analysis with Partially Paired Data
- Brier-UDAM for Paired Prediction-Rule Comparisons: Latent-Risk Interpretation, Observable Contrasts, and Inference
- Functional Inference for the Mean Curve of Intraday Spot Correlation
BibTeX
@misc{baumeister2026resamplingbased,
title = {{Resampling-Based Inference for Incomplete Paired Functional Data}},
author = {Marléne Baumeister and Lubna Amro and Łukasz Smaga and Markus Pauly},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.32777},
eprint = {2610.32777},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.32777},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Resampling-Based Inference for Incomplete Paired Functional Data
AU - Marléne Baumeister
AU - Lubna Amro
AU - Łukasz Smaga
AU - Markus Pauly
PY - 2026
DA - 2026/10/23
PB - arXiv
DO - 10.48550/arxiv.2610.32777
UR - https://arxiv.org/abs/2610.32777
AB - We consider testing equality of mean functions for paired functional data, where one entire function is missing. To use all available information from the paired and unpaired observations, we propose globalized test statistics that are based on a weighted sum of point-wise paired-t and Welch-t-type statistics. By the use of empirical process theory, we derive the asymptotic limit distributions of the resulting test statistics under the null hypothesis and show consistency against fixed alternatives. As they are non-pivotal and contain complex dependencies from the paired and incomplete observations, a permutation and parametric bootstrap procedures is proposed to compute critical values. The work includes a theoretical investigation of the asymptotic behaviour of the two resampling approaches and state regularity conditions under which the corresponding resampling tests are valid. Simulation studies show that the proposed methods provide accurate type I error control and competitive power for most settings under study. An illustrative analysis of functional emission data demonstrates the applicability of the methods.
ER -