Databases Reference
In-Depth Information
Figure 2.23
PCP ObsNo Removed
between other adjacent attributes. The crossing pattern that you see between
Weight and Ridge indicates an inverse relationship. Low values for Ridge
correspond to high values for Weight and vice versa. Direct relationships would
be seen as line segments that are close to parallel. There are no strong direct
relationships in this dataset.
One shortcoming of parallel plots is that you only visually see the relation-
ships between adjacent attributes. To see all of the relationships, you would
need to systematically drag each of the axes to be adjacent to each of the other
axes, pausing as you drag to observe the connection pattern between the axes.
For example, suppose your current axis sequence is: Crack, Density, Nub,
Ridge, andWeight and you would like to see all the relationships between Crack
and the other attributes.
Drag Crack to the right until it is situated between Density and Nub. Pause
to assess the relationship between Crack and Density to the left of the
Crack axis and Crack and Nub on the right. Next drag Crack to the right
again until it is situated between Ridge and Weight. Pause again to assess
the relationships.
Now rotate the plot all the way around (180 degrees) to look at the back
side (Figure 2.24). When you see a pattern similar to the patterns between
Density, Nub and Ridge they are indicative of independence (lack of
correlation) between the attributes.
