__/ [Charles Sweeney] on Friday 14 October 2005 19:21 \__
> Adam Risser wrote
>> Thanks Roy,
>> You seem correct that I will not be able to do trace-paths by
>> logs. I found this article interesting.
>> Maybe this is the route I should be taking (ala something browser
>> driven like hitbox)
> In my opinion, the route you should be taking is away from stats and
> more towards your visitors/users.
> You can't get anything useful from stats when compared to spending the
> same time getting feedback from your users, then making any changes
> based on that.
> If you deal with real visitors and give them what they want, then the
> stats will take care of themself.
I don't fully agree, Charles. You would not get statistically-valid facts
from one-to-one sessions with users. I do research into statistics, so I
can tell you it is not quite valid an argument. Better track the logs than
sit there and watch a user works on a PC of yours.
Some months ago I attended a meeting of the usability professionals in the
north-west. They actually do all of that user tracking in physical
environments. When I asked the speaker why she did not use stats packages
instead, she revealed her ignorance as to what stats packages can do.
Also, another meeting I attended was organised by the Statistical Society.
They invited Vapnik, a very famous machine learning figure and the inventor
of support vector machines. The head of our society had serious arguments
of frustration with Vapnik at the end of the session. Vapnik's ideas, much
like anything in machine learning, makes statistician a redundancy. This
extends beyond the field of statistics, but it's all an issue for another
day. It's neither the time nor the venue to discuss the demise in the need
for human intelligence, repetitive work in particular.
Roy S. Schestowitz | "World ends in five minutes - please log out"
http://Schestowitz.com | SuSE Linux | PGP-Key: 74572E8E
4:15am up 50 days 16:29, 3 users, load average: 0.08, 0.30, 0.46
http://iuron.com - next generation of search paradigms