Reproducing the analyses
The AdaReFun repository contains the package and simulation and data-analysis scripts. Follow the replication README for setup, saved-result checks, report generation, and numerical reruns. It records the package versions used in the study, including FMCP 0.2.0 and the regional fitting code.
Input dictionary
| Input | Meaning |
|---|---|
| y | Numeric scalar response, one per curve |
| x | Numeric matrix: rows are observations, columns are common argument locations |
| argvals | Strictly increasing locations in physical or declared normalized units |
| validationx | Independent response/vector and curve/matrix with the same grid |
| lambda / kappa | Explicit penalty and roughness tuning grids |
Training centering is performed by the package; scaling beyond this is an analysis choice. Missing or infinite values are rejected, not imputed.
Article datasets
Truck: 108 distributed rows, 90 acceleration measurements on seconds
1–90, particulate-emission response, distributed with PFLR 1.1.0. No
vehicle grouping identifier is available. NeoSpectra: 1,974 soil-sample
IDs and 601 reflectance measurements on 1350–2550 nm, response
log(1 + SOC percent). Repeat spectra were averaged within
sample before analysis. Their data sources and licenses remain separate
from the AdaReFun software license.
Seeds and preprocessing are specified in the run scripts. Folds and comparison predictions can be generated locally, or imported from an existing study after checking the recorded file hashes. The run guide describes both routes; no data or prediction files are bundled in the repository.