NCJ Number
146644
Journal
Journal of Forensic Sciences Volume: 39 Issue: 1 Dated: (January 1994) Pages: 64-73
Date Published
1994
Length
10 pages
Annotation
The likelihood ratio (LR) approach to DNA typing in criminal cases uses both size and pattern discrepancies between the crime scene profile of fragment lengths and the suspect profile for quantifying the strength of evidence.
Abstract
The LR indicates evidence is stronger when a pair of profiles matches closely than when a pair barely meets statistical matching criteria. The LR also indicates evidence is stronger when a pair of profiles exhibits differences in the same sign for alleles than when a pair exhibits differences in opposite sign. The LR differs from the match binning approach in which size and pattern discrepancies are considered by experts in the match decision but match closeness is not quantified. In contrast to match binning, the LR avoids an initial decision about whether two profiles match. For DNA typing in criminal cases, the LR numerator is the likelihood of a model that assumes suspect and crime sample fragment length profiles are identical. The LR denominator is the likelihood of a model that does not assume identical profiles. LR's for pairs of profiles that meet published statistical criteria for matching show a wide range of values, including some that indicate evidence is strongly against identity. The theoretical foundation of LR's is somewhat complex for jurors, but their numerical values can be labeled with broad descriptions, such as "very strong evidence of identity." Descriptions should be followed by reports on the proportion of times such strong results have been obtained from two different persons. Effects on the LR value of choices for target populations and assumptions about measurement error distribution should also be provided. 24 references, 2 tables, and 1 figure