On Sat, 6 Nov 2010 18:48:42 +0000, kay
wrote:
[color=blue][i]

;904444 Wrote:
On Fri, 5 Nov 2010 15:24:44 +0000, BTO GBW
wrote:

But only in participating gardens.


But a very large sample (around 10,000) [


Sure, but only in participating gardens.

-
in relation to
surrounding habitat, within garden practices and geographic location.-
Please tell us how you work that out?

The survey collects the relevant information.

What do you regard as "!relevant"?



-

Although there is variation in the ability of individual participants,
and in the amount of time they spend carrying out the recording, our
statistical models typically include a site-effect, which enables us to
control for this variation. From a statistical perspective, the sheer
size of the project increases its robustness, since it is the
underlying
patterns that are important.
-

Have you ever tried to count sparrows coming and going in a garden?

We have loads of them and I defy anyone to count them and know whether
the same ones are counted over and over again?


Counting is on the basis of the maximum number of birds seen at one time
(which is a lower bound on the actual number of birds).

Which gives no indication of the actual number of birds.

But many
analyses are simply looking at presence or absence of a species, and
actual numbers are irrelevant to this.


But only in participating gardens. That cannot be extended to whole
populations throughout the country, for the reasons I have given.

If you were doing a survey of car ownership you couldn't restrict
yourself to the most prosperous or most socially deprived areas to get
an accurate result.