Graphics Reference
In-Depth Information
Figure
.
.
Two
-D data projections. he two top plots are the data projections, and the two bottom
plots are illustrations of the projection coe
cients
Figure
.
.
Two
-D projections of
-D data
What Structure Is Interesting?
2.1.2
When we use tours, what are we looking for in the data? We search for data pro-
jections that are not bell-shaped and, hence, not normally distributed, for example,
clusters of points, outliers, nonlinear relationships, and low-dimensional substruc-
tures. All of these can be present in multivariate data but hidden from the viewer
who only chooses a few static projections. Figures
.
and
.
show some examples.
In Fig.
.
a scatterplot matrix of all pairwise plots is shown at let, and a tour
projectionisshownatright.hepairwiseplotsshowsomelinearassociationbetween
three variables, particularly between the variables TEMP and PRESS,and TEMP and
