Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing

What this work gives Aellic

Control of the false discovery rate across a family of tests. Necessary because the insight engine computes ~10 habits × ~12 metrics ≈ 120 tests — of which roughly 6 would break the p < 0.05 threshold by chance alone and be shown to the user as advice. A note on grading: `strength = review` is a makeshift here — this is an original methodological work with proof and simulation, for which the registry's vocabulary holds no value of its own. No PMID; the paper predates PubMed coverage of this journal.

Where this sits in the engine

1 declared site — constants and thresholds in the code that point at this entry.

• Behaviour Statistical safeguards 1 site