11.07.2015 Views

2DkcTXceO

2DkcTXceO

2DkcTXceO

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

112 Science mixes it up with statisticsthe chance of matching these alleles by chance is very low, then paternity isinferred.My work with plant paternity was about sex, but it could not be consideredsexy. I enjoyed this project very much, but it was of interest to a specializedcommunity. It was my next project that attracted the interest of other statisticians.In the late 1980s and early 1990s, DNA forensic inference was in itsinfancy. One of the earliest uses of this technique occurred in England. Basedon circumstantial evidence a boy was accused of the rape and murder of twogirls in his village. During his interrogation he asked for a blood test. As luckwould have it, Alec Jeffries, who had just developed a method of DNA fingerprinting,was located just six miles from the village. The boy’s blood was testedand he was found to be innocent. This was done by comparing the observedalleles at several genetic locations between the DNA left at the crime sceneto the DNA of the suspect. The choice of genetic locations was made so thatthere were a large number of possible alleles at each location. Consequentlythe probability of two people matching by chance was extremely low.This method of garnering evidence was tremendously powerful and as suchwas highly controversial. Genetic evidence had been used in paternity casesfor quite some time, but the impact of using DNA to convict people of seriouscrimes was much more compelling and there was an obvious need for seriousstatistical inquiry. Very early on, our colleague, Professor Neil Risch, wasinvited to work on the problem by a private company, LIFECODES, and alsoby the FBI. Neil invited us into collaboration. This was the beginning of anexciting period of investigation. The three of us published several papers andwere occasionally quoted in the New York Times. Although the setting andthe technology were new, many statistical questions were familiar from thepaternity project. Although we did not set out with a plan to get involved inthis hot topic, we would never have had this opportunity if we hadn’t gainedexpertise in the original botanical project.There were many aspects to this controversy. Here I’ll discuss one issue— the suitability of available reference databases for calculating the probabilityof a match between a suspect and the perpetrator. The question at issuewas how much people vary in their allele frequencies across populations. Forinstance, if an Asian reference sample is available, will it be suitable if the suspectis Korean? Naturally, controversy always rages most keenly when thereare little data available from which to definitively answer the questions. Let’sexamine this one guided by Sewell Wright. If we divide the world up intopopulations (continental groups) and subpopulations (ethnic groups withinacontinent),thenwecanbegintoexaminethisquestion.Wecanpartitionthe variance into various levels, among populations, among subpopulationswithin a population, among individuals within a subpopulation, and finally,layered over all of this is sampling error. Moderate sized samples were immediatelyavailable at the population level and it was apparent that populationsdid not differ strongly. But at the subpopulation level there was considerablesampling error, making it impossible to determine empirically if subpopu-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!