Data, reproducibility and validation

Input dictionary

Input	Meaning	Required separation
X_selection, y_selection	Numeric predictor matrix and response used to select a core	Independent of analysis observations
X_analysis, y_analysis	Numeric design and response used to estimate and test	Matching predictor names and order
Mandatory core	Predictor identities fixed in advance	Not selected using analysis responses
Test design and response	Untouched evaluation observations	Never used to choose tuning parameters
bootstrap_seed	Reproducible conditional-calibration seed	Record it with the draw count and version

Rows are observations; columns are predictors. Existing fit routines perform their documented centering and RMS scaling internally and return coefficients on the original scale. Do not add undocumented cross-sample preprocessing.

Reproduce an example, not a production claim

The quick-start script and saved figure belong to one simulated dataset. The package also retains historical benchmark sources under inst/benchmarks. Their presence is not a statement that every study was rerun for this site.

For an analysis record, save the call, seeds, package version, data-source identifiers, selection/analysis/test split and convergence/calibration output. Record sessionInfo() alongside results. Do not overwrite historical outputs.

Citation and contribution

Use citation("HDMaxShrink") for software and methods records, and cite the preprint as a preprint. Report bugs with a small synthetic example and sessionInfo(); do not attach private study data. The website build renders documentation and existing figures. It does not run production calibration, Monte Carlo studies or performance benchmarks.