Re: gnome-about-me

[John Williams <john williams lists gmail com>]

On Tue, 2005-08-02 at 16:11 -0700, Rob Adams wrote:

> I agree wholeheartedly that you're wasting your time.
Sigh.

> If you'd care to explain how this feature will enable us to estimate the 
> number of gnome users worldwide I would be very happy to hear it.
Sure.  As I said, it will establish a lower bound.  We can then compare
this with other estimates, see which is largest, and trust that one.


>   The 
> crucial bit of data, percentage of users who enable the checkbox, will 
> be unavailable.
Check,

>   Of course it will give you a lower bound, but so will 
> any number of methods already discussed.
Check.

> If you want to use the popularity-contest information to make an 
> estimate which is just as bad as the one we could make with this 
> "feature":  take percentage of debian popularity-contest users who have 
> GNOME installed.
OK.

>   Multiply by percentage of debian users with 
> popularity-contest installed.
I didn't realise that number was available.

>   Multiply by debian market share.  
Or this one.

> Multiply by linux desktop market share.
Or this one.

>   Multiply by number of internet 
> users.
OK, that is available.

> Note that in this chain, all the numbers can actually be estimated, 
> unlike the crucial "percentage of users with the count me feature 
> enabled".
Check.  But there are three numbers above that have to be estimate.
Multiplying them give multiplicative error expansion, even if they can
be validly estimated.  Can anyone help here?

>   The second one (percentage of debian users with 
> popularity-contest) could be estimated by analyzing server logs on 
> debian update servers when a new version of the popularity-contest 
> package is released,
Which pre-supposes knowledge of the total number of Debian users.  This
is huge news to me!  What is that number, and how was it generated and
validated?

>  and track the percentage of users performing 
> updates who also download the new package (this isn't perfect, but its 
> at least a good approximation, though there is likely bias in the 
> frequency of user updates and liklihood to have popularity-contest 
> installed).
Again, pre-supposes knowledge of total global Debian installations.

> The other numbers (the market share numbers) are available from any 
> number of research firms
For money, which we don't got.

> , and all have their own problems in determining 
> it.
This is the more important point.  They can give us a number, but can we
trust it?

>   And of course the total number of internet users is at best a 
> guess, but we can leave out that step and just use a percentage.
I would have thought that this number is the easiest of the lot to
estimate.

So, I'm dying to know, Rob:  what is the total number of Debian
installations worldwide, and of those, what is the total number of
Debian users?

And when you give me your answers, please tell me how I can trust their
validity?

-- 
John Williams
Research Analyst
Department of Marketing, Otago University
http://www.commerce.otago.ac.nz/marketing/staff/williamsj.html