Inherently at Risk: Low floral morphological variability predicts conservation risk

Reproducible analysis pipeline, derived data, and generated figures accompanying the article Inherently at Risk: Low floral morphological variability predicts conservation risk (Cambridge Prisms: Extinction; manuscript EXT-2026-0015). The study evaluates whether reduced within-species floral morphological variability — species-level coefficients of variation (CV) for seven floral dimensions — is associated with elevated conservation risk in Caribbean orchids (n = 240 species). Core pipeline: class-weighted logistic regression across m = 100 multiple imputations (MICE), complete-subset BIC model selection over all 127 predictor subsets, Monte Carlo cross-validation (N = 500), and a 10,000-permutation test of discrimination. Robustness analyses added during peer review, each addressing a specific reviewer concern: (i) phylogenetic non-independence — the Fritz & Purvis D statistic on an Open Tree of Life phylogeny, plus the full 127-model re-selection under phylogenetic logistic regression and two mixed-model frameworks; (ii) confounding by geographic range — a single-island-endemism covariate, under which the perianth-variability signal survives and strengthens; (iii) predictor collinearity — Spearman rank correlations and variance inflation factors; (iv) missing data — a full missingness analysis with robustness confirmed across three imputation strategies; and (v) trait synergy — targeted interaction models among the leading traits. Source code is licensed MIT (see LICENSE); data and figures are licensed CC-BY-4.0 (see LICENSE-DATA.md). Raw floral measurements are derived from Ackerman (2014), Orchid Flora of the Greater Antilles; the original flora is not redistributed here.