Jeffreys Scale
Definition
A logarithmic scale for Bayes factors proposed by Harold Jeffreys in Theory of Probability (1961). Levels run from 'barely worth mentioning' at log10(BF) between 0 and 0.5, through 'substantial', 'strong', and 'very strong', to 'decisive' at log10(BF) greater than 2. Widely referenced in forensic science but rarely applied without modification.
- Proposed by
- Harold Jeffreys
- Source
- Theory of Probability (1961)
- Metric used
- log10 of the Bayes factor
- Top category
- Decisive, log10(BF) greater than 2
- Common use
- Forensic likelihood-ratio verbal scales
Common questions
How does the Jeffreys scale differ from verbal scales used in forensic labs, such as ENFSI's?+
Forensic verbal scales are usually adapted from Jeffreys' original bands rather than copied directly, with different thresholds and wording chosen to fit the type of evidence being reported.
Why do critics warn against applying the Jeffreys scale directly in court?+
It was designed as a general statistical convention, not for forensic evidence evaluation, so translating a Bayes factor into a verbal label for a jury risks implying more precision or certainty than the underlying data supports.
Related terms
- ENFSI Guideline
- The European Network of Forensic Science Institutes Guideline for Evaluative Reporting in Forensic Science (2015, updated 2016). It defines the propositions framework,...
- Likelihood Ratio (LR)
- A ratio comparing the probability of findings under one stated proposition with their probability under another, conditional on relevant information. Its direction...
- Log10(LR)
- The base-10 logarithm of the likelihood ratio, sometimes called the weight of evidence (following I.J. Good). A log10(LR) of 1 corresponds to...
- Transposition Fallacy
- The error of treating the probability of the evidence given a hypothesis as though it were the probability of the hypothesis given...
- Verbal Equivalence Scale
- A table that assigns a verbal phrase to a range of LR values or log10(LR) values. The phrase is used in written...