Yes, I can see how the article MULTIDIMENSIONAL DETECTIVE helps a great deal in using parallel coordinates. The explanation of the first data set in that article is particularly cogent. Even that data set's exploration is very dependent on being able to select particular observations out and discovering changes in the patterns. Maybe that selection ability is more important than the rotatable 3D graphing I like so much. Of course the DataDesk example does not employ selectivity on its parallel coordinates graphic -- maybe because of the relatively low dimensionality of the sample data set there.
The later examples are less informative because there is no discussion of the methodology used for fitting the least squares model to the data and then creating the boundaries. Then I really get lost in their discussion of boundary points and exterior points. But their discussion of the first example was great and I really appreciate your reference to that paper. On Wed, Nov 13, 2013 at 1:15 AM, June Kim (김창준) <[email protected]> wrote: > Thanks for the interesting read. > > On the contrary to what's written on that page, Parallel Coordinates is my > favorite multi-dimensional exploratory data visualization method(For > categorical variables, I like mosaic plots). > > I would recommend reading Inselberg's seminal paper: Multidimensional > Detective at > http://www.cs.ucdavis.edu/~ma/ECS289H/papers/Inselberg1997.pdf > > -- (B=) ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
