Why can a hypothesis not take evidence from another test?

Because a test chosen after seeing the data is how a search turns into a finding by accident. The test and its fixed settings are registered first, and evidence that does not match them is refused.

Does a supported hypothesis mean something new?

No. It is supported by this data. Whether it is already known is a separate check: search Europe PMC from the hypothesis, read the papers it lists, and mark the one that reports it.

Is anything uploaded?

No. The ledger and every run it draws on stay in the browser, in your project.

Analyze

Hypothesis ledger: pre-register a hypothesis, attach runs, decide with e-BH

Register a hypothesis and the test that will answer it before any evidence exists, attach the runs that test it, and decide across every hypothesis with the false discovery rate controlled under any dependence. Evidence from another test, another fixed setting or data already counted is refused and the reason shown. The ledger lives in the project, every step is in its timeline, and it exports as one audit trail that imports back to the same verdicts.

Open Hypothesis ledger: pre-register a hypothesis, attach runs, decide with e-BH Open the Hypothesis ledger sample

Three restriction-site hypotheses pre-registered on eleven real genomes, decided by the site-avoidance runs attached to them

What it does

01Hypotheses registered with the test, its fixed settings and its direction before any evidence exists
02Evidence from another test, another fixed setting or data already counted is refused, with the reason shown
03κ-calibrated e-values, combined per hypothesis, decided by e-BH across every hypothesis under any dependence
04Every registration and every piece of evidence in the project timeline
05One JSON audit trail that exports and imports back to the same verdicts

What it runs on

The calibrator and e-BH are checked against POPPER's own implementation and, through e-BH's identity with Benjamini–Hochberg, against R's p.adjust. The panel runs on @ports/discover, @ports/enzymes, @ports/workflow.

Everything runs client-side. Files are parsed in the browser and projects are saved on this machine, so an unpublished sequence is never uploaded.

Questions

Why can a hypothesis not take evidence from another test?: Because a test chosen after seeing the data is how a search turns into a finding by accident. The test and its fixed settings are registered first, and evidence that does not match them is refused.
Does a supported hypothesis mean something new?: No. It is supported by this data. Whether it is already known is a separate check: search Europe PMC from the hypothesis, read the papers it lists, and mark the one that reports it.
Is anything uploaded?: No. The ledger and every run it draws on stay in the browser, in your project.

Also searched as

pre-registration · pre-registered hypotheses · e-values · e-BH · false discovery rate across hypotheses · hypothesis testing audit trail · p-hacking