Hello,
I'm a student currently studying Multi Dimensional Scaling thechniques.
I have difficulties finding answers to some of my questions. Therefore I
thought maybe some of you could help me out.
These are are my questions:
* While fitting the INDSCAL model, the proximity data (the raw data) are
first transformed to scalar products. I wonder why this is done. I could
only find that Carroll and Chang decided to do this because this makes
fitting the model more easy because this why the INDSCAL model can be
seen a special case of the CANDCOMP model. Does anybody know whether
there is anymore to this or could somebody explain this a little
further? Why are the scalar products being called "quasi scalar"?
* In the wandering ideal point model, one represents the subjects by
heuristic variables, while the objects are represented by fixed
variables. Why is this? Is this just a matter of convention? Could one
as well represent the subjects as static variables and the objects as
heuristic variables?
Many Thanks,
Dieter
--
Homepage: http://boa-web.org/
|