Ray Thomas wrote, about a confidence interval quote on the ONS website:
> The 'guide to accuracy' claim is difficult to defend. It is possible to
> believe that 'the ONC results remain the best central estimates' as the
> website says later on. But it is evident that the main source of error is
> deficiency of coverage. To talk about confidence levels based on
sampling
> error is to parade statistical method in a way that can impress only the
> unintelligent and the statistically naive.
I thought and hoped that sort of thing had long ago died out. When I worked
in local government, over 20 years ago, we had to carry out a survey of
industrial waste quantities, as a requirement of central government. The
whole thing was quite tightly specified in terms of sample sizes and so on.
But we were sent detailed, very clear instructions by Department of the
Environment statisticians on how to work out confidence intervals, based
(solely) on sampling error, for the amounts of waste of different types.
Since for some types of waste we had pretty useless sampling frames, and
since for non-hazardous, non-valuable waste the best you could find out was
that a company got rid of a skip-full every couple of weeks or so,
converting it all to tonnes and then producing a confidence interval by the
method we'd been instructed to use was a complete nonsense. In my local
authority there were (luckily) several competent statisticians (and me); we
complained and ignored the official instructions. I have used this as an
example of how not to apply statistical theory ever since. But I had no idea
it was still happening.
Kevin McConway
Department of Statistics, The Open university
******************************************************
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]
*******************************************************
|