Over the past couple of months I have sent two messages to
allstat suggesting that statisticians should show interest
publicly in situations where considerations about simple
probabilistic reasoning are mishandled with strong effects
for individuals. Specifically, I have mentioned the relative
frequencies which need to be considered on receiving a test
result that one is HIV+ (*), and the judicial deliberations
over the unfortunate Sally Clark who lost two children in
circumstances which each appeared to be those of cot death.
A third, and more greatly worrisome situation which has been
concerning me has been the way in which, in criminal investigations,
an enormously extreme surprise value is placed upon a DNA match, and
the related pressure from some quarters to establish a 100% DNA
database for forensic purposes.
In this latter concern, I find that I have been overtaken by a
radio program, broadcast yesterday (04-Jul-00); I heard most but
not all of the program. This was on BBC Radio 4, in the series
'File on Four'. The programme dealt particularly with the number of
codes possible in a particular type of DNA test (different tests deal
with different 'numbers of points' of the DNA code sequence),
and with whether code sequences from different individuals should be
regarded as completely independent realisations of a statistical
process (I paraphrase). It also dealt with non-statistical matters :
the handling mistakes which can be made in applying DNA tests (and
also fingerprint tests), the existence of considered unpublished
reportage which would tend to undermine confidence in the uniqueness
of a DNA test match outcome in practical forensic testing situations,
and at least one case of successful appeal against a conviction
based upon a DNA match where other forms of evidence showed
that the alleged perpetrator could not have done the deed.
File on Four is usually repeated on Sundays at 5 p.m., so I will
be listening carefully on 09-July. If you find this stuff engaging
and important then you too might wish to try and catch the repeat.
PS: I wonder what proportion of allstat subscribers are
Radio 4 listeners (but this is *not* a survey and *not* for reply to
the list !).
Norman Marsh
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|