The full workflow behind the catalog, running in your browser — your data never leaves it. Every component explained in Learn.
Long-format monitoring export: taxon, accession, storage code (Initial / −18C / 5C / Ambient), counts, and either YearsDifference or CollectionDate + TestStartDate. Columns are matched by name, any order.
15 tests parsed. Your data never leaves the browser — the entire computation is client-side. The engine reproduces the published catalog's own analysis at 92–99% field agreement on the real 29,020-test dataset it was built from (see Method for the residual sources).
The pipeline will: set an early-window baseline per accession → threshold at 70% of it → find each accession × storage's latest passing year → attach GLM + Wilson support → roll up to species intervals with supporting/potential counts → flag freeze sensitivity → assemble the publication sheet in censoring-honest notation.
