On Thu, 22 Apr 2004 21:09:50 +0100, Dave Fawthrop wrote:
> On Thu, 22 Apr 2004 19:31:25 +0100, Rev Adrian Kennard <`@o.gg> wrote:
>
> | Dave Fawthrop wrote:
> | >...
> | > If one looks at the month by month figures themselves, these are internally
> | > consistent and only show wild swings, as one could reasonably expect, at
> | > Christmas. Were ISPs to be telling outright lies in answer to the
> | > questions in the surveys wild swings would occur. I therefore conclude
> | > that ISPs, as Adrian suggests, may well not give *totally* correct answers,
> | > but generally do the best they can. IMO the surveys give generally good
> | > results, but are not as accurate as statistical theory may suggest.
> | >
> | > It is clearly a Good Idea that the Government, and others, know roughly
> | > what is happening in specific markets to facilitate planning. Perhaps
> | > Adrian cold suggest anothher method of collecting this data.
> |
> | I am not sure surveys, as such, are a bad idea. In practice it would be
> | better to try and collect stats that an ISP can easily automate, and
> | perhaps collect the stats in an automated way (if we have to have thsi
> | in the first place). It would also help to collect stats in a way that
> | avoids duplication in the first place.
> |
> | The problem with the survey (which, incidentally they don't include us
> | in any more!) was that it asked questions which often made no sense,
> | contradicted themselves, and were not the most appropriate of questions
> | anyway. They also seemed to lack any concept of normal statistical
> | methods (and I only did O-level stats).
> |
> | They failed to define, even when challenged, what they meant by terms
> | like "dialup" and "permanent connection". They failed completely to
> | sensibly define "virtual ISP" which was a term key to the whole survey.
> |
> | We made no differenciation as to whether someone is residential or
> | business, but they had no way to take stats that were not broken down as
> | residential or business. They said to "estimate" the ratio but would not
> | accept my estimate of 50%+/-50%! They also failed to understand that
> | someone may use a service for both, and indeed the terms "residential"
> | and "business" are not mutually exclusive or fully inclusive anyway.
> |
> | Apparently return of the survey was a legal requirement, but they failed
> | to answer if completion of the survey was a legal requirement (the
> | wording was odd which made be suspicious that the requirement may simply
> | be to return it!).
> |
> | There were many other stupidities in the whole thing. I am concerned
> | that if the survey was typical of their stats collection then their
> | results may be dubious as best in other areas too.
> |
> | All IMHO, of course.
>
> Computing people and political people look at the world in different ways.
> Computing people expect questions to be specific and to have exact answers.
> Political people hope to glean some useful indications, and are probably
> aware that the results produced are inaccurate.
Except the these figure could be collated accurately, and it does not
require talking to any ISP's whatsoever. All you need to do is trawl the
billing databases of BT, Telewest and NTL. For example BT know how many
premises have telephone lines, which lines are business or private, which
lines are dialing up to the internet, which are ISDN, and which are ADSL
enabled, on an exchange by exchange basis. Telewest and NTL hold similar
information as well.
JAB.
--
Jonathan A. Buzzard Email: jonathan (at) buzzard.me.uk
Northumberland, United Kingdom. Tel: +44 1661-832195
|
Similar Threads
Linear Mode
