SWGDAM panel urges numeric likelihood ratios, offers single verbal scale and guidance on exclusions

At a webinar hosted by RTI International—s Forensic Technology Center of Excellence, members of an ad hoc Scientific Working Group on DNA Analysis Methods (SWGDAM) laid out recommendations for reporting likelihood ratios (LRs) in forensic DNA analysis, emphasizing numeric transparency and consistent language for nontechnical users.

"The numerical value of the LR is the critical piece of information, and that must be reported," said Sean Monpetit, a SWGDAM member and San Diego Police Department DNA technical manager. SWGDAM—s guidance, presenters said, treats a verbal predicate as optional and always subordinate to the numeric LR.

The working group recommended a single, standardized verbal scale to accompany numeric LRs so that analysts, attorneys and juries have consistent interpretive context. As Monpetit explained, the group favored a streamlined scale with categories such as "limited," "moderate," "strong" and "very strong" support tied to LR ranges informed by empirical data. The full verbal scale should appear alongside the numeric LR in reports and testimony so recipients can see the mapping between numbers and verbal descriptors.

Panelists also addressed how laboratories should report when multiple LR values are calculated using different population allele-frequency settings. The group concluded that reporting a single value is acceptable but recommended that the reported value be the lowest of the calculated LRs so that reports err on the side of the person of interest. "In general... the lowest value will be the one that is most beneficial to the person of interest," the group said.

On exclusions and "inconclusive" labels, SWGDAM recommended a conservative approach. The working group suggested laboratories may adopt a numeric threshold below which they would declare an outright exclusion, and it recommended 0.01 (an LR of 1/100) as an upper bound for that practice while allowing labs to set their own policy. Conversely, the group recommended against using broad "inconclusive" zones that span ranges of LRs around 1. Sean Monpetit said such zones can create unintended consequences and do not prevent adventitious (chance) support for an inclusionary proposition.

Dr. Tamara Moretti of the FBI laboratory summarized the empirical basis for these recommendations. The dataset the committee discussed included internal validation mixtures from 31 laboratories (2,825 mixtures) and a later STRmix reanalysis that produced roughly 28,000,000 non-contributor comparisons. Moretti quoted external commentary that "Adventitious support is in fact expected" in very large non-contributor comparisons, adding that most adventitious-support results are concentrated at lower LR ranges in those datasets.

The panelists emphasized clear statement of the propositions being compared (for example, the evidence is compared under the hypothesis that the victim and the person of interest were contributors versus an alternate pairing), and they warned against phrasing that treats an LR as a probability that a particular person is the donor. Stephen Myers (California Department of Justice) advised analysts to define proposition pairings explicitly before using simplified wording in testimony so that jurors and other listeners do not conflate LRs with posterior probabilities.

During a lengthy Q&A, audience members asked whether the reanalyses included systems other than STRmix; presenters said the large multi-lab reanalysis that informed much of the empirical discussion was STRmix-based, though committee membership included users of other systems. Panelists also said several agencies (including CalDOJ and the FBI) have adopted the recommendations in whole or in part, while some agencies (OCME was cited) are not incorporating all recommendations.

The webinar closed with the host noting that a PDF of the slide deck and a recording will be made available through RTI—s Forensic Technology Center of Excellence resources and that presenters provided contact information in the chat for follow-up questions.

The working group—s recommendations and supporting materials (including the STRmix reanalysis and cited literature) are available on the SWGDAM website and in the referenced publications cited during the webinar. The webinar presenters encouraged laboratories to document the LR calculation method, to provide context for any single reported LR, and to avoid substituting verbal predicates for numeric transparency.