ResultAtlas explains statistical results with attention to both interpretation and provenance: what a result means, where a commonly repeated rule came from, and whether that rule applies universally.
Source hierarchy
For production-critical claims, ResultAtlas prefers sources that are closest to the method or implementation being described:
- Original methodological works, standards, and authoritative technical documentation.
- Strong methodological reviews, peer-reviewed research, and official institutional guidance.
- Secondary explanations only when they add context that stronger sources do not supply, and only within the evidence they actually support.
A source’s existence is not treated as proof of a claim. When only an abstract, metadata record, or later review is available, the wording is limited to what that evidence can support.
How thresholds are labeled
ResultAtlas separates different kinds of statistical rules instead of presenting every number as a universal cutoff.
| Label | What it means |
|---|---|
| Formal | Defined by a mathematical, legal, or standards-based rule in the stated setting. |
| Original convention | A convention traced to an identifiable original source. |
| Discipline convention | A practice used within a particular field or decision context. |
| Software default | A value or rule implemented by a specific software tool. |
| Rule of thumb | A practical heuristic, not a law or guaranteed decision boundary. |
| Disputed | A rule whose interpretation, attribution, or usefulness is materially contested. |
| No universal threshold | Available evidence does not support one context-free cutoff. |
These labels are not a ranking from best to worst. They tell the reader what kind of authority a rule has and how much context it needs.
Interpretation standard
Each guide is organized around a specific reader question. It aims to:
- give a direct interpretation early;
- distinguish odds, risks, effects, uncertainty, fit, and diagnostics accurately;
- explain important limitations without burying the answer;
- keep software behavior separate from general methodology;
- identify when the same value can mean different things in different settings.
ResultAtlas does not add thresholds from memory, merge unlike conventions into a single scale, or use citations merely to make a page look authoritative.
Health and professional boundaries
Health-related examples are used to explain statistical methods. They are not clinical advice, treatment guidance, diagnosis, or an assessment of an individual person’s risk. Statistical interpretation also does not replace the domain expertise, study design review, or decision criteria required in a professional setting.
Corrections and updates
Guides may be updated when a source locator is corrected, a stronger primary source becomes available, software behavior changes, or wording could overstate the evidence. A correction should preserve the page’s defined interpretation job and make the source trail clearer.
If a primary source cannot be independently verified, ResultAtlas may rely on a clearly identified later methodological source and state that limitation in reader-facing terms. Unsupported numbers and attributions are omitted rather than filled in from memory.