Volume 1 Number 1 - September 2003

Statistical Retorts: The Next Generation

Years ago, Correlation Research published a newsletter for attorneys that focused on the use and misuse of statistical evidence in litigation. Dubbed Statistical Retorts, the publication was time-consuming and expensive to write, edit, print and mail out. While this newsletter generated many nice comments and even a little business, eventually, the cost and effort involved proved prohibitive. I recall lamenting: "If only there were a more efficient and less expensive way to disseminate this kind of information....."

Fast forward 15 years. Voila, the Internet! So I recently thought: "Why not resurrect our old print newsletter as a shiny new e-newsletter?" Its continuing mission: to promote statistical awareness and sophistication among attorneys. We hope you will find this modern edition as informative, and even entertaining, as the classic version.

Sincerely,

Herb Weisberg

President of Correlation Research

Manacles on 34th Street?

Statistical evidence often plays a critical role in discrimination cases, but a recent New York case broke some new ground. On June 2, in Federal district court of lower Manhattan, a class-action suit was filed by Sharon Simmons-Thomas against Macy's and its parent company, Federated Department Stores. Ms. Simmons-Thomas during last Christmas season was allegedly detained, wrongfully accused of shoplifting, deprived of legally purchased property, and coerced into paying several hundred dollars to Macy's.

Her suit claims that a virtual campaign of false accusations and intimidation, including threats of detainment (sometimes using handcuffs or confinement) aimed at minority shoppers has been undertaken throughout the Macy's chain. The class-action is being pursued by the New York law firm of Thompson Wigdor and Gilly, which has produced statistical evidence in support of their contentions. In particular, it is alleged that in the Herald Square store, out of 1,600 individuals apprehended for shoplifting in the year 2002, only 8% were white. A fascinating case in which statistics should play a major role in proving or countering the charges!

The Prosecutor's Fallacy

A recently much-publicized statistical issue is the so-called Prosecutor's Fallacy. The problem typically occurs when a prosecutor has collected some evidence of a rare occurrence that is consistent with the defendant's guilt. While such an occurrence can provide evidence against the defendant, the strength of this evidence is sometimes grossly exaggerated because of the Prosecutor's Fallacy.

As an example, suppose that a particular defendant's DNA matches that found at the scene of a crime. This DNA pattern occurs in 1 out of 1,000 in the general population. Joe Defendant is found to have the DNA pattern. Does this imply that Joe is very likely to be the perpetrator? Not necessarily. To assume so uncritically would be an example of the Prosecutor's Fallacy.

To understand why, let us consider two different ways that Joe might have been singled out as a suspect. First, suppose that an eye witness to the crime identified the perpetrator's general physical characteristics. Joe was picked up because he was found in the neighborhood and generally fit the description. Imagine further that these characteristics might also fit 5,000 other men residing in the general area. Among these 5,000 we would expect around 5 to have a DNA match. So, the probability would be only about 17% (1 out of 6) that Joe is guilty.

On the other hand, what if Joe were collared based on a very precise description given by a credible witness? Then, there might be only a handful of men who would be possible suspects. In this case, the DNA evidence, coupled with the precise description would lead to a very high probability that Joe is the culprit. For example, if the witness testimony alone might lead a jury to ascribe a likelihood of around 20% that Joe is the guilty party, then adding the DNA evidence would increase the probability to 95%.

The take-home message is that the evidence provided by a rare occurrence must be evaluated in conjunction with the other information that is available. Confusion about this principle and its application caused sufficient concern recently to prompt a letter by Peter Green, the president of Britain's Royal Statistical Society (www.rss.org.uk) admonishing the Lord Chancellor "to take steps to ensure that statistical evidence is presented only by appropriately qualified statistical experts, as would be the case for any other form of expert evidence." We would add that the expert should also be one who is able to explain statistical concepts clearly for a non-technical decision-maker.

Herbert I. Weisberg, Ph.D.
61 Pheasant Landing Road
Needham, MA 02492-1000
Phone (781) 455-6850
Fax (781) 444-9563

email: hweisberg@correlation.com

Please let us know what you think about our newsletter. Thank you.