Critical Role of Statistics in Development and Validation of Forensic Methods

Saturday, February 16, 2013
Room 310 (Hynes Convention Center)
Karen Kafadar , Indiana University, Bloomington, IN
Statistics has played a key role in the research, development,
and validation of DNA forensics, as well as in the inferences
(conclusions) obtained from forensic evidence in general. Statisticians
also have been important contributors to many areas of science, such as
chemistry (chemometrics), biology (genomics), medicine (clinical
trials), engineering (imaging) and agriculture (crop yield), leading
to valuable advances that extend to multiple fields (spectral analysis,
penalized regression, sequential analysis, pattern recognition,
experimental design).  Yet the involvement of statistics in forensic
science more generally (beyond DNA) has been surprisingly low, given
the value it has demonstrated thus far (e.g., assessment of bullet
lead evidence; significance of findings in the U.S. anthrax investigations)
and its contributions to advances in other branches of science.  In
this talk, we first discuss two cases (bullet lead, 2004; anthrax, 2011)
in which statistics played a vital role in the findings.  We then
suggest other forensic disciplines where research is needed, how
statistical methods can contribute, and ways in which statisticians
can participate, to achieve reliable forensic evidence consistent
with the scientific method (e.g., low error rates of false positives and
false negatives) and raise the level of confidence in the forensic system.


1. Role of statistics in science (chemistry, biology,
medicine, agriculture, computer science, etc.) -- 3-4 slides

2. Success of Statistical methods in previous forensic cases

a. DNA analysis (1-2 slides) (1992?)
b. Bullet lead (3-4 slides)  (2003)
c. Anthrax (2 slides)        (2011)

3. Conventional role of statistics: expert testimony (1-2 pages;
cite cases in which statisticians have testified)

4. A more important role for statistics: Greater involvement at
the front end (development of well-tuned methods, rather than
simply at the end when called in for prosecution or defense).
Examples: pattern recognition (specifically fingerprints),
experimental designs of validation and confirmation.

5. Statisticians: A path forward (How we can achieve greater involvement)