Examples
Worked examples
- Is an instance
Choosing the outcome variable scale (raw, log, ranked) after seeing the data.
- Is an instance
Choosing the inclusion window for a time-series analysis after pilot analyses.
Counter-examples
Looks similar, but isn't
- Not an instance
A fully pre-specified analysis with no choices made after data inspection.
- Not an instance
A registered report that locks the analysis plan.
Editorial commentary
Gelman and Loken (2014) named the 'garden of forking paths' to describe the implicit branching that occurs when analyses are conducted after looking at the data. Even without explicit p-hacking, the latent multiverse of analyses an analyst could have run produces inferential bias. The concept motivates multiverse analysis and specification curves as corrective tools.
References
- Gelman, Loken, 'The garden of forking paths' (Columbia University working paper, 2013/2014).
Also known as
forking-paths problem
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Forking paths"
vocab-term-identifier="https://casrai.org/dictionary/term/forking-paths" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"name": "Forking paths",
"identifier": "https://casrai.org/dictionary/term/forking-paths",
"description": "The phenomenon by which the cumulative effect of many small, data-contingent analytical choices inflates false-positive rates even when each individual choice appears defensible.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/reproducibility-and-computational-research/",
"url": "https://casrai.org/dictionary/term/forking-paths",
"sameAs": [
"forking-paths problem"
],
"license": "https://creativecommons.org/licenses/by/4.0/"
}
