Neutrino Atlas
EN
  1. Terms
  2. Analysis & measurement
  3. Statistics & inference
  4. Statistical significance (σ)
  5. p-value

p-value

ConceptLevel 4 of 4

The probability, assuming only background, of obtaining a result at least as extreme as the one observed.

A small p-value indicates tension with the background-only (null) hypothesis, but it is not the probability that the hypothesis is true. It is usually translated into σ.

In plain words

Assume, for the sake of argument, that nothing new is going on and the detector sees only background. How often would chance alone produce a result at least as striking as the one observed? That fraction is the p-value. A tiny p-value says the data sit uncomfortably with the 'nothing new' assumption; it does not say how likely the new idea is.

In detail

p = P(t ≥ tobs | H₀)t: a test statistic such as an event count or a likelihood ratio; tobs: its observed value; H₀: the null (background-only) hypothesis.

The p-value depends on the choice of test statistic and on whether one or both tails count as 'extreme'. Particle physicists usually quote one-sided p-values for excesses and convert them into σ. When the test statistic follows a known distribution, as Wilks' theorem guarantees for likelihood ratios in large samples, p can be computed directly; otherwise it is found from simulated pseudo-experiments.

Common misreadings

  • It is not the probability that the background-only hypothesis is true.
  • It is not the probability that the result is a fluke given that it was seen.
  • A small p-value does not mean a large or important effect; with enough data, tiny deviations become significant.
  • A p-value from one place in a wide search must be corrected for the look-elsewhere effect.
  • It says nothing about systematic errors that the background model does not include.

Example

LSND reported in 2001 an excess of about 88 electron-antineutrino events above background, a p-value corresponding to 3.8σ. The p-value was small, but whether the excess means a sterile neutrino or an unmodelled background has been debated for more than two decades.

Frequently asked questions

What p-value counts as significant?

There is no universal line. Particle physics uses about 1.3 × 10⁻³ (3σ) for evidence and 2.9 × 10⁻⁷ (5σ) for discovery, far stricter than the 0.05 common in other sciences.

Is a large p-value proof that there is no signal?

No. It only means the data are compatible with background. A weak signal may still be present, which is why experiments also quote upper limits.

Original papers

  1. TheoryG. Cowan, K. Cranmer, E. Gross and O. Vitells, Asymptotic formulae for likelihood-based tests of new physics, Eur. Phys. J. C 71, 1554 (2011)DOI 10.1140/epjc/s10052-011-1554-0arXiv:1007.1727

Mentioned in 3 pages

Last updated: · Editorial standard

Cite this page

Text

Neutrino Atlas (2026). “p-value”. Retrieved 22 November 2026, from https://neutrino-atlas.com/terms/p-value/

BibTeX

@misc{neutrinoatlas2026termspvalue,
  title        = {p-value},
  author       = {{Neutrino Atlas}},
  year         = {2026},
  url          = {https://neutrino-atlas.com/terms/p-value/},
  note         = {Accessed 2026-11-22},
  language     = {english}
}