For researchers
Name the method, check the data, keep the evidence.
You analyse data for a living: a trial statistician, a postdoc, a clinical scientist, or a group that wants to repeat a result somebody else published.
What you get
The method in your table, named and versioned
The platform reads the paper, works out which analysis was used and how it was parameterised, and binds that to a numbered operation. The same paper gives the same answer on every run.
Your data checked against the paper before anything is run
Whether the source data is reachable, under what licence, with which accession is recorded first. Where a paper names no route to its data, that is the finding, and the platform says so rather than proceeding quietly.
Results that carry their own provenance
Input hash, declared parameters, the result, and hashes again over the whole thing, in a bundle that can be handed to a reviewer as it stands.
What is true today
- Group comparison, by one-way analysis of variance with Welch pairwise tests under Holm's correction, and incidence-rate modelling are implemented and proved against synthetic data with known answers.
- The corpus names 24 further analysis operations. They are ranked by how many papers each one unlocks and written in that order.