2007/2/25, Ray Saintonge <saintonge(a)telus.net>et>:
I was going to suggest a different sport (perhaps American football) as
the other group. A possible distortion from randomizing in one sport is
that the person who adds the free pictures is likely to do so in batches
that are accumulated independently of our randomizing process.
I think that comparing two sports will give much more distortion, exactly
because of the batches in that case are NOT independent of our randomizing
process. If you use randomizing within one group, such a batch of pictures
would increase the numbers for both groups, and probably rightly so. If you
use a different sports as your check group, it means that such a batch would
go all to one side, thus causing quite a bit of distortion which may well
not be related to the difference we are looking at.
Another problem I see is that it might take quite a long time to get
results. I doubt whether a year would be enough to get a meaningful
difference, even if a large number of pages are involved.
--
Andre Engels, andreengels(a)gmail.com
ICQ: 6260644 -- Skype: a_engels