Graphics Reference
In-Depth Information
Figure . . Distance-based linking. he view on the right shows a scatterplot of the two major fielding
statistics. he user has clicked at approx. the ( , . ) coordinate to investigate this cluster and the degree
of interest is set to reflect the closeness to this point in this plot. he let view shows a hex binning of
the players' physical statistics, and the mean degree of interest for a bin is coded using the same
brightness coding as the other plot. In both cases we are using a brightness scale with black mapping to
% selected (complete interest) and light gray mapping to % selected (no interest)
data - defining a location in the data space based on the view and then using dis-
tance as measured in the data space. Normalization of the data dimensions would
also probably be necessary for this. he transfer function that converts distance into
degree of interest is also important. he choice of function involves similar issues as
are found with choice of kernel and bandwidth for kernel density estimation. he
function should give a value of one at a distance of zero and should decrease to zero
as the distance increases, so that the point selected is given complete interest and
points further away are assigned decreasing degrees of interest. Following are some
choicesthathavebeenfoundeffective inpractice.Noresearchhasbeendoneintothe
optimality of any of the following possibilities; they have been chosen in an ad hoc
manner. he third method was used in the creation of Fig. . above.
DoI = max( ,
d
C), for some constant C.
DoI =
d
(d
+
)
DoI =
(d
+
C), for some constant C.
Linking from Multiple Views
9.6
InSect. . wediscussedhowtocombine anexisting degree-of-interestmeasurewith
a new selection, in the case of a binary degree of interest. he goal was to allow users
to repeatedly modify the degree of interest by successive interactions with the same
view. he same principle can be used to link multiple views together. We term this
Search WWH ::




Custom Search