Madam, - Toby Joyce (March 14th) makes no attempt to answer either the statistical or commonsense objections to the Lancet report of 655,000 Iraqi dead made by myself and Sean Coleman. Instead, he simply asserts that the debate has been settled in the Lancet's favour.
The credibility of the Lancet group's methods must stand or fall by comparison of its 2004 report of 98,000 dead with that of the Iraqi Living Conditions Survey of 24,000 in the same year. The methodology which Toby Joyce extols as "a tried and trusted technique", that of cluster sampling, was used by both, but to far greater effect by the latter.
As stated previously, the ILCS sampled 22 times as many households as the Lancet's 990, and its teams spent a median time of 83 minutes per household asking a range of questions including that concerning mortality. The Lancet's teams, on its own admission, could each complete a cluster of 40 households in one day. Even a gruelling working day of 10 hours would thus allow only 15 minutes per household, allowing for neither breaks nor travelling time between supposedly randomly distributed houses. But then, as the report tells us, "decisions on sampling sites were made by the field manager", an admission which casts serious doubt on the randomness of the whole procedure.
Given the slender sampling base of the Lancet group's methods, it is no wonder that its 2004 figure came with a 95 per cent confidence interval of 8,000 to 194,000. Contrary to Toby Joyce's definition of "confidence interval", this means that there is a 95 per cent probability of the true figure lying between these two extremes (just as if pollsters tell us that 40 per cent of voters intend to vote for Fianna Fáil, with a 95 per cent confidence interval of 37 to 43 per cent, we understand that the true level of support is very likely to be somewhere between these limits). It is reasonable to use the colloquial term "margin of error", as I did, to substitute for "95 per cent confidence interval", without being guilty of "confusing" the two. And margins of error as wide as those in the two Lancet reports convey only one message: "Go back and do a lot more sampling".
If Mr Joyce were not aware of the weak statistical basis for the Lancet's conclusions, why bother, as he does, to excuse the small number of clusters in the sample (as necessary to lessen the risk to the interviewers)? Mr Joyce regards the Lancet as a "prestigious" scientific journal, perhaps unaware of the damage done to its reputation in the recent past by the campaigning agenda of its editor, Dr Richard Horton. In February 1998 the latter caused a worldwide scare by rushing into print with the notorious article alleging a causative link between the MMR vaccine and autism based on claims made by parents of only eight children.
As to his acting "without fear or favour", that is hardly the impression to be gained from his far-left harangue at the "Time To Go" rally in Manchester last September, in which he accused the "axis of Anglo-American imperialism" of initiating and sustaining the sectarian violence in Iraq - oh, and of spreading "global death" with its "policies of hate and oblivion". - Yours,
DERMOT MELEADY, Clontarf, Dublin 3.