[Sigia-l] Cluster analysis is dead - was Jakob wrong?

[Eric Scheid]

On 8/10/04 11:50 AM, "Trent Mankelow" <trent.mankelow at optimalusability.com>
wrote:

> If the modes are fundamentally opposed, such as year vs
> genre, then the kind of average you would get out of cluster analysis
> with all the sorts included could be a non representative average, ie
> junk.

looking at the EzSort output, isn't that 'confidence of agreement'
represented quantitatively ?

maybe this problem could be solved by looking at the hierarchy problem ...
if the system was able to handle the user's explicit card sort hierarchies,
maybe the system could also synthesize implicit hierarchies (eg. from a
population of users split on genre vs year, it could reveal that top level
hierarchy).

I'm also curious as to how you intend to handle the Buddha Nature of items.
This also has implications on split modes.

BTW, if you're writing up some stats software, I wouldn¹t mind a simplified
tool for inputting a table of data and having it spit out which columns are
more significant than others. This would be a real boon for writing up
personas from survey data.

e.