Description
Pure-R Core Engine for Optimal Data Analysis (ODA / MultiODA).
Description
Pure-R implementation of univariate binary-class ODA (UniODA), univariate multiclass ODA (MultiODA), and binary Classification Tree Analysis (CTA). Supports ordered and categorical attributes, priors-on inverse-frequency weighting, MAXSENS / SAMPLEREP / first-identified tie-breaking, true leave-one-out cross-validation, and Monte Carlo Fisher-randomization p-values. Covered UniODA, MultiODA, and binary CTA fixtures are tested for parity against MegaODA.exe and CTA.exe outputs.
README.md
Fixture provenance
This directory contains canonical test fixtures used for oda parity and regression testing. Fixtures are validated against MegaODA.exe / CTA.exe golden outputs.
myeloma/
Public survminer/GEO fixture, transformed for CTA parity tests.
- Source:
myelomadataset from the survminer R package (GEO ID GSE4581). - Public data; no PHI; no private institutional data.
- Used for weighted CTA, MINDENOM (1/30/56), LOO STABLE, missing-code handling, endpoint counts, and no-tree behavior.
- See
myeloma/README.mdfor full provenance.
cta_demo/
Synthetic demonstration fixture distributed with CTA.exe.
- Used for unweighted CTA parity (MINDENOM = 1 and 8).
- No confidential data.
vignettes/
Vignette article fixtures (Example-1 through Example-4 from MPE Chapter 2 and 4).
- These are published-example datasets from Yarnold & Soltysik (2005/2016).
- Used for ODA/MultiODA parity tests (UniODA, categorical, directional).
Protein fixture (Chapter 3 multiclass example)
Protein fixture provenance: derived from a previously published dataset; publication PMID 6643432. Used only for fixture parity/regression testing.