Sounds easy. I wonder why this "study" doesn't mention a p value. The grant
must not have been large enough to fund someone with any experience using R,
or god forbid, a pencil.
On Thu, May 7, 2009 at 10:32 PM, Robert Rohde <rarohde(a)gmail.com> wrote:
On Thu, May 7, 2009 at 8:58 PM, Brian
<Brian.Mingus(a)colorado.edu> wrote:
Quite frankly the advice that you should only use
five subjects makes no
sense. The appeal to Nielsen's authority is not going to work on me or
anyone else who understands why the scientific method exists. It's
unscientific thinking and it's going cause to you waste money. You're
going
to draw conclusions based on results that simply
aren't valid, and you
won't
know it until the study is over and you
didn't make progress.
Careful analysis of site data could allow you to draw some conclusions.
I'm
curious how you're planning to go about that.
Dependent/independent
variables?
If five subjects, chosen at random, all have the same problem, then
with 95% confidence you can predict that at least half of the
population will report having this problem.
This kind of work generally focuses on BIG problems, and you don't
need a huge sample to identify some of the most common issues. In
things like UI development it would be surprising if there weren't
complaints reported by most of the subjects. You may overlook some
other problems, but when coming up with a list of common problems to
work on, I would say that 15 subjects is plenty.
-Robert Rohde
_______________________________________________
foundation-l mailing list
foundation-l(a)lists.wikimedia.org
Unsubscribe:
https://lists.wikimedia.org/mailman/listinfo/foundation-l