Estimation and inference in penalized function-on-function regression with within-curve residual dependence

Fabian Scheipl · Department of Statistics, LMU Munich · Munich Center for Machine Learning (MCML)

Penalized function-on-function regression, as implemented in the R package refund, considers all grid values as independent conditional on the linear predictor, but some dependence typically remains. We study the consequences for estimates and pointwise intervals in a simulation and a plasmode study on nine real datasets. Under dependence, restricted maximum likelihood (REML) undersmooths the coefficient surface: in our simulations' main settings, its relative error is two to four times that of curve-blocked neighborhood cross-validation (NCV). In our plasmode study, both fits have large errors, possibly through weak identification, and NCV is usually, but not always, better. Under dependence, the independence-model-based intervals for means and surfaces undercover severely. We propose and evaluate a curve-clustered sandwich covariance with a leverage adjustment for the penalized hat matrix with Satterthwaite degrees of freedom which brings coverage close to nominal in most settings. Its intervals for function-on-function effects are wide because REML's estimate is noisy. The same construction at the NCV estimate is narrow but undercovers. We therefore recommend fitting twice: NCV for the shape of the surface, REML with these intervals for inference. For estimated means, intercept functions and scalar covariate effects, one REML fit with these intervals usually suffices.

Reproducing the results

The paper, the supplement and the detailed results render from the aggregate files in the code repository, without the raw records. To recompute those aggregates, download the task records you need from the Zenodo deposit into the folder raw/ of your clone, where the scripts look for them:

git clone https://github.com/fabian-s/fof-inf.git
cd fof-inf
bash scripts/get-raw.sh list                          # the available parts
bash scripts/get-raw.sh results r4-results study.rds  # for example, the main study

The script downloads each archive, extracts it to raw/<name>/, checks every record against its checksum and removes the archive. All parts together take about 86 GB once extracted, so fetch only the ones you need; the README says which part belongs to which analysis. Without the script:

mkdir -p raw
curl -L -o raw/results.tar.xz "https://zenodo.org/records/23916388/files/results.tar.xz?download=1"
tar -xJf raw/results.tar.xz -C raw

Records of the two cardiology datasets are patient and volunteer data and are not in the deposit; the repository holds only aggregate results for them, so those results cannot be recomputed.

Citation

Scheipl, F. (2026). Estimation and inference in penalized function-on-function regression
with within-curve residual dependence. arXiv:2610.04677.
Raw task records: https://doi.org/10.5281/zenodo.23916388.

Code: GPL-3. Paper, supplement, results and raw task records: CC BY 4.0.