Home Messages Index
[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index

Re: W3Counter Global Web Stats December 17, 2006

  • Subject: Re: W3Counter Global Web Stats December 17, 2006
  • From: Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx>
  • Date: Thu, 28 Dec 2006 21:02:22 +0000
  • Newsgroups: comp.os.linux.advocacy
  • Organization: schestowitz.com / Netscape
  • References: <5YidnWTNRtq_lQnYnZ2dnUVZ_r_inZ2d@speakeasy.net> <pan.2006.12.28.19.57.00.214739@ncoldns.com>
  • Reply-to: newsgroups@xxxxxxxxxxxxxxx
  • User-agent: KNode/0.7.2
__/ [ Kelsey Bjarnason ] on Thursday 28 December 2006 19:57 \__

> On Thu, 28 Dec 2006 10:08:37 -0800, John Bailo wrote:
> 
>> 
>> http://www.w3counter.com/globalstats/
> 
> Hmm...
> 
> "This report shows statistics aggregated across all active websites
> tracked by W3Counter. It was last generated on December 17, 2006  based on
> an analysis of 6,974,802  distinct visits across 4,018 websites."
> 
> "4,018 web sites" is pretty damned insignificant.  As is 7 million visits;
> it wouldn't surprise me much to find that, say, microsoft.com got that
> many in a day.  Where's the breakdown of _which_ sites are tracked this
> way?  The breakdown showing that they're not tracking msn.com and
> microsoft.com, thus wildly skewing the stats?
> 
> Of course, there's other issues with this as well.  Take, say,
> www.download.com.  They have a moderate collection of Linux software.
> However, I rarely go there, because virtually everything I need or want is
> in the repositories - unless I'm using Windows, in which case, I'll go
> there.  Result?  They're going to show something like 100:1 usage of
> Windows over Linux, even for me, despite the reality being pretty much the
> exact opposite.
> 
> Web stats are completely meaningless unless one can demonstrate both that
> the sites in question are equally interesting to similar proportions of
> Windows and Linux users (e.g. the site appeals to about half of the users
> of each OS), that the sites drive the users of each OS to make comparable
> numbers of visits, and that the sites don't themselves impose a bias
> (msn.com is not going to get a hell of a lot of Linux users heading to it.)
> 
> Of course one then also has to remove the biases from proxying, caching
> and the like, account for exactly how much browser ident changing is going
> on, etc.
> 
> Lastly, the coverage has to be complete enough to be meaningful; 4,018
> sites and 7 million visits doesn't seem to mean much unless that's an "in
> the last five minutes" report.  Even if it is, collect the data for a
> month, let's see the spread over, say, a million sites, after removing the
> biases and skews.
> 
> You do have those figures, right?  I mean, you wouldn't be posting this
> stuff here knowing, going in, that it's worthless, so you must have
> already done all those steps, right?  So why post the raw data, why not
> post the meaningful results?

http://en.wikipedia.org/wiki/Lies%2C_Damn_Lies%2C_and_Statistics

-- 
                        ~~ Kind greetings and happy holidays!

Roy S. Schestowitz      | Anonymous posters are more frequently disregarded
http://Schestowitz.com  |  RHAT GNU/Linux   ¦     PGP-Key: 0x74572E8E
  9:00pm  up 71 days  7:14,  6 users,  load average: 1.04, 0.81, 0.77
      http://iuron.com - help build a non-profit search engine

[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index