Myra Samuels Memorial Lecture

A Statistician in Court

Professor Bruce S. Weir
William Neal Reynolds Professor of Statistics and Genetics
North Carolina State University

Start Date and Time: Wed, 23 Apr 1997, 4:30 PM

End Date and Time: Wed, 23 Apr 1997, 5:30 PM

Venue: North Ballroom of the Purdue Memorial Union

Refreshments: A reception will be held from 5:30-6:30 PM in the North Ballroom of the Purdue Memorial Union.

Abstract:

When a match is found between DNA profiles in an evidentiary stain and a blood sample from a suspect, it is usual to supply some number expressing the strength of this evidence. Because DNA profiling is still relatively new, it is also usual for both sides in a court case to call upon statisticians to justify or criticize these numbers.

There is growing agreement among statisticians that simply quoting the frequency of the matching profile in some population is not satisfactory, and can even be misleading. It is better to provide a likelihood ratio, formed as the probability of the profile under the prosecution explanation divided by the probability of the profile under the defense explanation. These likelihood ratios depend on population genetic principles, and do not incorporate non-quantifiable effects that might be caused by fraud or incompetence.

Likelihood ratios are especially necessary when an evidentiary stain contains the DNA profile of more than one contributor. The evidence can be substantially over- or under-stated if only the frequencies with which unknown contributors have the stain profile are presented in court.

Although statisticians have played a role in interpreting DNA profile matches, including those in the Simpson case, the highly discriminatory nature of the profiles now being determined suggests that this need may have passed.