And here is the paradox: Governments and businesses still employ
statisticians because they know statistics produces the information they
need to make informed decisions.
Ted Harding
<Ted.Harding@NESSIE To: [log in to unmask]
.MCC.AC.UK> cc: (bcc: David Mcnulty/GB/QUES/ICI)
Sent by: email list Subject: Re: Cluster analysis
for Radical
Statistics
<[log in to unmask]
AC.UK>
09/11/2004 16:27
Please respond to
Ted Harding
Well, I thought that Ray Thomas's reply was itself in the
same spirit as John Taylor's original message.
Statisticians, it seems from messages to this list, are only
slowly coming to realise the truth about Statistics.
Let's start a little way back. It is well established in Law
that Statistics is unreliable, deceptive and misleading: there
are seminal judicial decisions to this effect. It could be said
that statistical evidence is for practical purposes now almost
inadmissible in a Court of Law. Don't forget that Jack Straw
is a lawyer, so we can trust him.
However, elsewhere it's open season on Statistics.
Honest statisticians admit, in their reports, the pitfalls of
their designs and analyses. The authors of the Lancet study on
Iraqi civilian deaths did just that, quite thoroughly it seems.
Thus do the grouse fly straight into the guns.
From Hansard (yesterday in the House of Lords):
Lord Lamont of Lerwick asked Her Majesty's Government:
What is their response to the claims in the Lancet magazine
that 100,000 civilians have died in Iraq as the result of
the military action by the coalition forces.
The Minister of State, Foreign and Commonwealth Office
(Baroness Symons of Vernham Dean):
My Lords, there are no reliable or comprehensive figures for
Iraqi civilian casualties. Such estimates that exist are not
comparable in terms of periods covered or methodologies used.
The Lancet article suggests a range of between 8,000 and
194,000 deaths over the period March 2003 to September 2004,
while the Iraq body count website suggests a range of just
over 14,270 to just over 16,400 over the same period.
The Iraqi Ministry of Health says that just under 4,000
civilians have been killed in the past six months. My right
honourable friend the Foreign Secretary has stated that when
the Lancet estimates have been analysed, he will make a
Statement to another place, and I shall put a copy in your
Lordships' Library.
Lord Lamont of Lerwick:
My Lords, I thank the Minister for that reply. Does she not
agree that if the upper end of the Lancet's estimates were
right, that would make nonsense of the claim that the type
of modern warfare waged by the coalition in Iraq was the most
humanitarian in history? Is it not ironic that the Government
are now defending what has happened by citing lower estimates
of casualties, which previously they rubbished? And is it not
regrettable that it has taken this article for the Foreign
Secretary to say that he will do what he should have done
long ago--that is, to make a proper estimate of the civilian
casualties in this war?
Baroness Symons of Vernham Dean:
My Lords, the fact that we have not had reliable figures has
been a matter of enormous concern to Her Majesty's Government
for quite some time. So I assure the noble Lord, Lord Lamont,
that this concern has not arisen only in the light of the
article in the Lancet. We take these estimated figures very
seriously, and the Foreign Secretary has made that clear.
But we are cautious about them because they are very different
from the figures emerging from other sources, and they have
been questioned by a number of independent observers.
I am sure that the noble Lord, Lord Lamont, will have read the
article in the Lancet, as, indeed, have I. I think that it
bears more detailed analysis, and that work is being undertaken
on the basis of the Lancet figures in order that my right
honourable friend can put forward a detailed Statement.
Thus the range 8,000-194,000 deaths from the Lancet study (which
was their confidence interval) becomes, along with the Iraq Body
Count range of 14,270 - 16,400 and the Iraqi Minister of Health's
statement of 4,000, to shoot down a study that wears its methodology
on its sleeve.
Lord Lamont astutely focussed on the top end of the CI to point out
that "if the upper end is right" [as in principle it could be]
then things looked pretty bad (and this is the first time I've
seen a reference to any claim that the Iraq War is "the most
humanitarian in history"). Which led to a "holding reply" which
dodges Lord Lamont's question; meanwhile the Statistics is made
to look ridiculous once again, with all sorts of different results
being wheeled out.
However, the most malicious assault on the dignity of Statistics
yesteday came from Tony Blair himself:
From Hansard (yesterday in the House of Commons):
In respect of the figure of 100,000 that has been bandied
about as the number of people who have died in Iraq over
the past period, I saw today, looking at the figures in
greater detail, that the figure is extrapolated from the
recorded deaths of 61 people. When one sees that, one can
understand how much propaganda is coming through on the issue.
Well, of course! To infer 100,000 deaths from 61! it must be
boloney! Mind you, that's quoted from Hansard; it's not exactly
what TB said [transcribed from the spoken word; source Radio4,
"Yesterday in Parliament"]:
"In respect of the figure in this figure of a hundred thousand
that has been bandied about about the numbers of people that
have died, ermm in Iraq err over that past period of time,
imean, err, when I actually see today, looking at these figures
in greater detail, this a hundred thousand figure is an
extrapolation from recorded death of 61 people, then I think one
can just understand quite how much propaganda is coming through
on this issue."
which goes to show what TB has gained from close proximity to GWB.
Clearly, therefore, Statistics is low on Street Cred, is suffering
from Disrespect.
Disrespect is one thing: it gives all comers a right to mix in
and throw Statistics around as brutally as they wish. Since
Statistics is clearly so contemptible, it is easy then, when the
results are incorrect, to simply correct them. For an example,
see my recent post about Drug Treatment Statistics, where the
incorrect figure of 126,000 produced by statisticians at Manchester
University was allegedly corrected by the the Dept of Health to
154,000 -- a figure which clearly was correct because that is
what it had to be. [By the way, no-one has responded to my
request for information this post.]
Statisticians, though, however brutally Statistics is treated,
and however much Disrespect it attracts, perhaps still see
themselves as custodians of the morality of Information Integrity,
from raw data through its various transformations under analysis
to the final conclusions (with caveats and all). So much for that:
If the posting "FW: We've Been Bad" from Jay Ginn is anything to
go by, it seems we can't even trust our raw data. If the data are
wrong, simply correct the data. If there is a possiblity, prior
to the event, that the data will be wrong, then make sure that¸
in the event, they will be right. Doesn't this make Statistics
irrelevant, except as an Aunt Sally for opportunists?
So where are our rights as statisticians?
Best wishes to all,
Ted.
Lots of comment cut.
______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email
______________________________________________________________________
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
*******************************************************
|