Paul Novitski wrote:
At 11:11 AM 12/6/04, Felix Miata wrote:
Fresh meat: http://www.useit.com/alertbox/20041206.html
Yes, but only 605 respondents?! Yikes, that's a small sample.
Nielsen's results, satisfying as they are to one allergic to
commercialism, would carry more weight if the sample size were
significantly greater. Perhaps someone blessed with a memory for
statistical math can confirm how large a website-viewing population can
be significantly sampled by just 605 respondents.
Responses from 1000 people, picked out / selected at random, is
calculated to give an error of +/- 3% for a larger group of more than
100 millions.
If you pick responders from the same, small, group over and over again,
the error will slowly rise towards a useless +/- 50%.
Even worse: if you get responses from a group of followers, then it is
always biased and useless.
One can always question any statistical results-- no matter how big a
sample. One can even use statistics to prove the reverse, if one like
to. Statistics based on samples are the most used and abused form of
manipulation there is.
Didn't see how those responses were filtered, but if there was any
serious balance (statistically speaking) then the error should be less
than +/- 10%. That would make a pretty strong case the way the numbers
came out, but no one need to believe it.
Statistics are almost as much fun to work with as xhtml and CSS...
...you can always get the result you want. :-)
Georg
******************************************************
The discussion list for http://webstandardsgroup.org/
See http://webstandardsgroup.org/mail/guidelines.cfm
for some hints on posting to the list & getting help
******************************************************
