Adding optima can destroy the identifiability of α: code and data

R scripts, derived CSV tables, figure sources and the sealed pre-registrationunderlying the manuscript "A closed-form decomposition of OUM likelihood gain:predicting the identifiability of α before optima are fitted" (SystematicBiology, submitted). Contents- scripts/: 23 R scripts in pipeline order; ou_fast.R is the self-contained OU likelihood engine used by all downstream scripts, and paper_figures_tables.R regenerates Figures 1–3 and Tables S1–S2 from scratch.- data/: 16 derived CSV tables and one RDS bundle (real_results.rds: palm tree + model-comparison objects read by most scripts).- figures/: figure sources as published (PNG 300 dpi + PDF vector).- preregistration/: the sealed pre-registration (preregistration_second_control.md, SHA-256 319e405d495ed89c...), its out-of-sample results, the verification report and the SHA-256 seal.- SHA256SUMS.txt: checksums for archive integrity. All analyses were run in R 4.6.1 with ape 5.8.1, phytools 2.5.2, OUwie 3.0.3and ouch 2.20. The archive reproduces every number, figure and table in themanuscript. No unpublished data are used. DOI: 10.5281/zenodo.<issued> (cite in both the data-availability statementand the reference list)