rolescry - Name-Blind Variable-Role Detection by Data Signature

Deterministic, name-blind detection of variable roles (group, outcome, survival or duration time and event, coupled binary-continuous pairs, calendar time index, paired and agreement measurements, repeated measures and panels, scale and rating items, subject or entity identifier, covariate) in tabular data from any domain -- clinical, financial, social, experimental. Roles are assigned from each column's information-theoretic signature -- Shannon entropy, normalized mutual information, and distributional shape -- rather than from column names, so renaming columns to 'col_1', 'col_2', ... does not change the result ("Data inspice, non nomen"). An optional, capped name-based hint and automatic header-row detection are also provided. No large language models and no external data transmission. Extracted from the 'MDStatR' biostatistics engine; see Boynukara (2026) <doi:10.5281/zenodo.20707791>.