ANULUM / DISCOVERY

An early warning must also survive the control

Understand why the early-warning study compares transition detection at a matched false-alarm budget across several data modalities.

Companion to an archived preprint

The question

At the same false-alarm allowance, do generic warning indicators identify transitions more often than no-transition controls?

The reported result

The study reports that its generic early-warning suite did not clear the statistical significance criterion across the tested modalities under the matched-false-alarm protocol. A visible trend in a selected record is a different question from reliable discrimination against controls.

Read the original publication ↗

How to read the experiment

  1. Declare transition and control records
  2. Match the false-alarm budget
  3. Compare detection with a permutation test
Study-design schematic, not a numerical figure. Read the paper for its quantitative results and uncertainty.

What the result does not establish

The conclusion is bounded to the tested detectors, corpora and evaluation protocol. It does not establish that every early-warning method is ineffective in every domain.

Follow the evidence

For reproduction, use the original paper’s protocol and repository instructions, including its data and hardware requirements. The browser learning demos explain concepts; they do not reproduce these experimental studies.

Discuss a reproduction