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

Look-elsewhere effect

ConceptLevel 4 of 4Also called: trials factor, post-trial significance

The inflation of apparent significance when one searches many places (energies, sky directions, parameters).

A local 4σ excess in an all-sky neutrino map may correspond to a much lower global significance once all trial directions are counted. Telescopes therefore quote both local and post-trial significances.

In plain words

Roll one die and get a six: unremarkable. Roll a hundred dice and find that one of them shows six five times in a row: also unremarkable, once you remember how many dice there were. Searches that scan many energies, sky directions or parameter values face the same trap. Somewhere a fluctuation will look striking, simply because so many places were inspected.

Local and global significance

p_global ≈ 1 − (1 − p_local)^N ≈ N · p_localN: effective number of independent places searched (the trials factor); valid for small p_local.

The local p-value refers to the most significant spot; the global, or post-trial, p-value asks how often a spot at least that significant would appear anywhere in the search. When the search regions overlap, as neighbouring sky pixels or adjacent mass values do, the effective N is smaller than the raw count and is usually found with pseudo-experiments. Eilam Gross and Ofer Vitells described a widely used shortcut in 2010.

Where it matters

  • All-sky searches by neutrino telescopes test hundreds of thousands of directions, so a local 4σ hotspot can be globally insignificant. IceCube quotes both values; its 4.2σ result for NGC 1068 already accounts for the 110 sources in its catalogue.
  • Sterile neutrino searches scan a grid of Δm² and mixing angles, and an oscillation-like wiggle can appear by chance at some Δm².

How experiments handle it

They fix the search region in advance, often as part of a blind analysis; they state the trials factor; and they treat a local excess as a target for an independent, pre-registered follow-up rather than as a result in itself.

Frequently asked questions

Does the look-elsewhere effect apply to a single fixed hypothesis?

No. If the location, energy or parameter value was chosen before looking, there is only one trial and the local significance is the right one.

Why not always quote the global value?

Both are useful. The local value says how unusual the spot is; the global value says how unusual it is to find such a spot at all. Papers report both so readers can judge.

Original papers

  1. TheoryE. Gross and O. Vitells, Trial factors for the look elsewhere effect in high energy physics, Eur. Phys. J. C 70, 525 (2010)DOI 10.1140/epjc/s10052-010-1470-8arXiv:1005.1891

Mentioned in 7 pages

Last updated: · Editorial standard

Cite this page

Text

Neutrino Atlas (2026). “Look-elsewhere effect”. Retrieved 22 November 2026, from https://neutrino-atlas.com/terms/look-elsewhere-effect/

BibTeX

@misc{neutrinoatlas2026termslookelsewhereeffect,
  title        = {Look-elsewhere effect},
  author       = {{Neutrino Atlas}},
  year         = {2026},
  url          = {https://neutrino-atlas.com/terms/look-elsewhere-effect/},
  note         = {Accessed 2026-11-22},
  language     = {english}
}