Examples
Worked examples
- Is an instance
A 1,000-specification curve of a wage-effect estimate across choices of controls, sample restrictions, and functional forms.
- Is an instance
A specification curve in a developmental psychology paper showing effect-size variation across 64 plausible models.
Counter-examples
Looks similar, but isn't
- Not an instance
A single sensitivity analysis varying one covariate.
- Not an instance
A pre-registered single model.
Editorial commentary
Specification curve analysis is a transparency technique that reports how an estimated effect changes across a large, pre-defined set of theoretically defensible model specifications, rather than reporting a single preferred estimate. Every specification — a particular combination of control variables, sample restrictions, outcome definitions, and functional forms — that a reasonable analyst could justify is run, and the resulting effect estimates are plotted in order of size, alongside an indicator of which specifications were statistically significant. An accompanying inferential test asks whether the observed pattern of estimates across the whole curve is more extreme than what would be expected under a specified null model, rather than relying on the significance of any one specification.
The purpose is to make researcher degrees of freedom visible instead of hidden. A single headline estimate gives a reader no way to judge how much of the result depends on discretionary analytic choices; a specification curve shows the full range those choices could have produced, so a reader can judge whether the finding is a general pattern or an artefact of one favourable specification.
What it is not
Specification curve analysis is not a statistics tutorial topic to be taught here, and it is not simply “trying a few robustness checks” — see robustness check for that narrower, more selective practice. It is also distinct from, though closely related to and often produced alongside, a multiverse analysis: a specification curve is centred on plotting and testing one estimated effect across specifications, typically varying analytic choices one dimension at a time in a fairly linear model-comparison framing, while a multiverse analysis (below) is framed around enumerating combinations of upstream data-processing decisions as well as modelling choices, and often reports the proportion of the resulting universe of analyses that cross a given threshold rather than a single ordered curve. In practice the two techniques overlap and are frequently cited together, but a page claiming to be about specification curves should not silently become a page about multiverse analysis, or vice versa.
For a research office or journal, the practical question specification curve analysis raises is not how to compute one, but what it signals: authors who publish one are disclosing, rather than concealing, the sensitivity of their result to analytic choices, which is precisely the transparency the TOP Guidelines' Analytic Methods Transparency standard is designed to encourage.
References
- Simonsohn, Simmons, Nelson, 'Specification curve analysis' (Nature Human Behaviour, 2020).
Also known as
specification curve analysis · SCA
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Specification curve"
vocab-term-identifier="https://casrai.org/dictionary/term/specification-curve" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"@id": "https://casrai.org/dictionary/term/specification-curve",
"name": "Specification curve",
"identifier": "https://casrai.org/dictionary/term/specification-curve",
"description": "An analytic and visual technique that plots the estimated effect across a large set of theoretically defensible model specifications, ordered by effect size, to convey the sensitivity of the result to analytical choices.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/reproducibility#set",
"url": "https://casrai.org/dictionary/term/specification-curve",
"sameAs": [
"specification curve analysis",
"SCA"
],
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@id": "https://casrai.org/#organization"
},
"author": {
"@id": "https://casrai.org/#editorial-team"
},
"datePublished": "2026-05-21T02:22:49",
"dateModified": "2026-08-22T15:41:28",
"inLanguage": "en-GB",
"isAccessibleForFree": true
}
