Not to worry, Ray -- the message from "John Taylor" is obviously a jest.
How could somebody post seriously to a statistical discussion group while
professing complete unawareness of the widely-used methods of cluster
sampling (and associated statistical analysis methods)?
By the way, there is a nice exposition of the statistical issues for lay,
nontechnical audiences (by an activist and educational researcher named
Steven Soldz) at
http://www.zmag.org/content/showarticle.cfm?SectionID=15&ItemID=6565
Soldz argues that the Lancet authors bent over backwards to be
conservative (in the statistical, not the partisan, sense, of course) by
ignoring the implicit stratification of the sample.
On Mon, 8 Nov 2004, Ray Thomas wrote:
> Geoff Hoon picked up this "extrapolation" idea from a Fellow of the RSS -
> his boss Jack Straw.
>
> Why should this bit of appraisal be limited to limited to 'cluster
> analysis'. Surely the extrapolation idea applies to all ONS social surveys
> that all use samples with most using stratified samples just like the Iraq
> study. So are statistics from any survey now dismissable among members of
> the government on the grounds that they are extrapolations.
>
> Will the ONS and RSS be communicating with Jack Straw and Geoff Hoon on what
> appears to be attack on a fundamental idea supporting the production of
> National Statistics?
>
> Ray Thomas
> 35 Passmore, Tinkers Bridge, Milton Keynes MK6 3DY
> Email: [log in to unmask]
> Tel 01908 679081 Fax 01908 550401
>
> ----- Original Message -----
> From: "Taylor, John G LIS" <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Monday, November 08, 2004 4:27 PM
> Subject: Cluster analysis
>
>> Did anyone else hear our esteemed Minister of Defence on Radio 4 on Friday
>> morning?
>>
>> Geoff "Has the war started?" Hoon may not know - or be told - much about
>> defence but he put John Humphrys right about the statistical analysis
>> suggesting an excess of 100,000 Iraqi deaths since Iraq was liberated by
>> Coalition forces.
>> He explained that the methodology was suspect. Apparently the method is
>> to
>> count deaths in one area and then "extrapolate" to the whole country.
>> Obviously if you pick somewhere like Fallujah and do this the figure you
>> obtain is going to be skewed.
>> I am surprised that the statisticians involved did not realise this flaw -
>> so obvious now that Mr Hoon has pointed it out - and I am a bit
>> disappointed
>> that no-one on this list subjected the original report - or presumably the
>> whole theory of cluster sampling - to such a searching critique either.
>> It is a great loss to statistics that, against all the odds, he kept his
>> current job after the Hutton and Butler reports butI hope that Mr Tony
>> continues not to trouble him overmuch with all that military stuff and he
>> has the time to continue to contribute to statistical theory in such an
>> incisive way.
> **********************************
>
******************************************************
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]
*******************************************************
|