Skip to content

Likelihood Ratio (LR)

Definition

The ratio of two conditional probabilities: the probability of the observed evidence given the prosecution's hypothesis (same source), divided by the probability of the observed evidence given the defence's hypothesis (different source). An LR greater than 1 supports the prosecution hypothesis; an LR less than 1 supports the defence hypothesis. The LR is not a probability of guilt.

Type
Bayesian statistical metric
Field
Forensic science (cross-disciplinary)
Governs
Evaluative reporting of evidence strength
Direction
LR > 1 supports prosecution; LR < 1 supports defence; LR = 1 is neutral
Key promoters
ENFSI, UK Forensic Science Regulator

Common questions

What is a likelihood ratio in forensic science?+

A likelihood ratio (LR) is the probability of the observed evidence under the prosecution's hypothesis divided by the probability of the same evidence under the defence's hypothesis. An LR greater than 1 means the evidence supports the prosecution. An LR less than 1 means it supports the defence. An LR of exactly 1 means the evidence is neutral.

Why does it matter that the LR is not a probability of guilt?+

The LR only measures the relative strength of the evidence. It tells the court how much more probable the evidence is under one hypothesis compared to the other. Deciding guilt requires combining that with everything else in the case, which is the jury's job, not the forensic expert's.

Where is the likelihood ratio used in forensic practice?+

It is used across many disciplines: DNA profiling, fingerprints, speaker comparison, document examination, tool marks, soil, paint, and forensic botany. Bodies like ENFSI and the UK Forensic Science Regulator promote LR-based reporting as a more principled alternative to older categorical conclusions like 'match' or 'no match'.

Related terms

Prosecutor's Fallacy
The error of treating the RMP (or its reciprocal) as the probability that the defendant is innocent, or as the probability that...
Random Match Probability (RMP)
The probability that a randomly chosen unrelated person from the relevant population would match the evidence profile by chance. A very small...
Daubert Standard
The US federal evidentiary standard (Daubert v. Merrell Dow Pharmaceuticals, 1993) requiring that expert testimony be based on scientifically valid methods with...
Defence Fallacy
The converse error of inflating the importance of the RMP by arguing that, because many people in the population share the profile,...
Evaluative Reporting
A framework for expressing forensic conclusions in terms of the probability of the evidence given the prosecution hypothesis versus the defence hypothesis...
Frye Standard
The US legal test for admissibility of scientific evidence, originating from Frye v. United States (1923), which required that a technique be...
Prior Odds
The ratio of the probability of the prosecution hypothesis to the probability of the defence hypothesis before the forensic evidence in question...
Bayes' Theorem
The mathematical rule that relates prior odds, the likelihood ratio, and posterior odds. In the odds form: O(Hp|E) = O(Hp) x LR....
Posterior Odds
The ratio of the probability of the prosecution hypothesis to the probability of the defence hypothesis after the forensic evidence is taken...
Product Rule
The rule that, under independence, the joint probability of a multi-locus genotype equals the product of the individual locus genotype frequencies. Used...
Theta (FST) Correction
A correction factor applied to allele frequency estimates to account for population sub-structure (the tendency for allele frequencies to differ between sub-populations)....
Verbal Scale
The ENFSI translation of numerical likelihood ratios into courtroom language: very strong support (LR over 10,000), strong support (LR 1,000-10,000), moderate support...

Explained in these topics

Your journey to becoming a forensic professional starts here.

Practice with mock tests, learn from structured notes, and get your questions answered by a global forensic community, all in one place.