Forensic Science Statistics
by Moderator on Aug.04, 2011, under Topics for discussion
Forensic DNA testing has seen the use of statistics in forensic science beginning with the start of DNA testing, only about 20 years ago. This is a new area for experts to educate lay juries in how to interpret what constitutes a “match” between two items of evidence.
How do we adapt our methods to a (possibly new kind of) statistical basis and (2) how do we best communicate statistical results to laypersons?
Yet, statistics can lead the inexperienced astray as this short lecture presented on the Internet on TED ® demonstrates. Using statistics to explain events is more difficult to understand then it first appears:
http://www.ted.com/talks/peter_donnelly_shows_how_stats_fool_juries.html
Some critics question the use of terms like unique, similar, and identical. Might statistics help laypersons to understand relationships between a known source and an evidence sample? For example, automobile paint collected at a hit and run matches a suspect vehicle. If a jury were told that this color paint is found in one in ten vehicles, would this be helpful information, rather then just stating, “the paint matches.”
What do you think the role of statistics is in trying to explain complex matters to lay juries should be? Does it help clarify or confuse matters?
August 12th, 2011 on 6:29 am
I think what we need is a new statistical approach entirely: DNA got to where it is because of the simple structure of the molecule, the predictability of the population statistics, and the selection of the loci for forensic purposes. If you think about it, until you multiply the allele frequencies together, DNA is still class level evidence–its power comes from the selection of the independent alleles.
When we get to other materials, either biological or manufactured, things become far more complicated, not the least of which is that components of manufactured are intentionally selected for their properties. New statisical models need to be explored that (1) reflect the nature, structure, and history of the evidence and (2) is relatively easy to explain to a jury. Things like epidemiology, meta-analysis, data-mining, economic theory (are transfers of evidence like economic exchanges?) should be considered. I’ve said this all before but no one seems to pay attention.
Are you listening, academic programs? Great masters’ thesis ideas…