(This page is under contruction.)
 

(material copied from Intro:)

This family of smooths has an interesting connection to "computer vision".  The theoretical side of this area of computer science involves mathematical modelling of human vision.  "Scale Space" is one of these models.  See Lindebergh (1994) Scale Space Theory in Computer Vision for an excellent introduction.  It turns out that scale space is a family of Gaussian kernel smooths, indexed by the bandwidth.  Thus the above two "family plots" are also "scale space plots".  The family of smooths models vision in the sense that large bandwidths correspond to stepping back and "looking macroscopically" only at coarse aspects of the big picture, while small bandwidths correspond to zooming in and "looking microscopically" at fine details.  Much deeper discussion of the scale space view of smoothing is given in Section D.
 

(enhancement of mode tree).

("spectrum" of smooths)

For more about SiZer, inquire by email from marron@stat.unc.edu.
 

Back to SiZer Main Page

Back to Data Analysis Table of Contents

Back to Marron's Home Page