I was quoting the statistics, I wasn't pretenting to be a statisitcian
In defence of Roy Meadow
Lancet, Vol 366, July 2,2005
Richard Horton (editor of Lancet)
Beyound reasonable doubt
Plus Magazine, 2002
Muliple sudden infant deaths--coincidence or beyound coincidence
Paediatric and Perinatal Epidemiology 2004, 18, 320-326
Sir Roy Meadow is a pediatrician who is well known for his research in child abuse. As this is written the UK General is considering the removal of Meadow from the medical registry for flawed statistical estimates that he made while testifying as an expert witness in a 1999 case in which a Sally Clark was convicted for murdering her two baby boys and given a lifetime sentence. In Chance News 11.01 we provided a letter that the Royal Statistical Society sent to the Lord Chancellor explaining the errors in Meadow's testimony and in Chance News 11.04 we mentioned a very nice article by Helen Joyce who discussed the statical issues in more detail. This article was based on unpublished articles written by Roy Hill professor of mathematics at the UK university of Salford. However Hill has recent published his analysis in his 2004 article mentioned above. We will use this article to give a review of the statistical issues.
The first death was recordered as a natural cot death. Å "cot death" or "crib death" are other names for SIDS (sudden infant death syndrome). Hill writes:
The definition of SIDS which is still current was formulated by the American pathologist Beckwitgh, in 1969 as follows:
the suddent death of a baby that is unexpected hy history and in whom a thorough post-mortem examination fails to demonstrate and adequatge cause of death.
A cot death is also called Sudden Infant Death Syndrome Death (sids). But after her second baby boy died she was tried and convicted of murdering both of her children and given a lifetime sentance.
Roy Meadow testified in her trial and stated that chance of two natural cot deaths was 1 in 73 million suggesting that two natural cot deaths was beyond belief. This estimate came from a study called the "Confidential Inquiry for Stillbirths and Deaths in Infancy" (CESDI). The study gave detailed information about the deaths of all babies in five regions of England between 1993 and 1996. The report of the study estimated that the chance of a cot death was 1 in 1,303. But if the child was from an afluint non-smoking family with the mother aged over 26, then the chance decreased to 1 in 8,543) Since Sally Clark was in the this group, Meadow assume the chance of the first child was a cot death was 1/8,543 and squaring this he obtained the 1 in 73 million estimate for the chance of two cot deaths.
Of cource, soon after the trial it was pointed out that this multiplication assumed independence of the two events which surely would not be the case. In fact Enviornment, biological and genetic factors would suggest a lower probability for the second child. Then it was pointed out that this was a classic case of the Prosecurers Paradox which in the case of DNA fingerprinting has been officially recognized by the courts in England.
To be continued