Databases Reference
In-Depth Information
plotted at a predefined height. All other density heights are plotted relative to
the maximum. The “Flatten” axis option is useful when you have an axis
with a dominant distribution value. For example, you may have a Yes/No
nominal attribute in which 90% of the observations are “Yes”. The density
height at the “Yes” location on the axis will be so much greater than all other
density points that it totally dominates and makes distribution assessments
along the other axes difficult. To prevent that domination, right-click on the
axis that dominates and select the “Flatten” option. That axis will be drawn
with density height encoding and the other axes will adjust their heights
relative to the next best height density point. It sounds complicated, but
remember, if you see a single point dominating all others with respect to
height density, making the others look flat, then flatten that axis and the
others will be allowed to grow.
Exercise 2.5
Use the VisMiner parallel coordinate plot to locate and extract sub-populations
from the out5d.csv dataset. This dataset contains satellite and sensor readings
collected from a 128
128 grid in Western Australia - SPOT a measurement
from a satellite image, magnetics, and three radiometric measures: potassium,
thorium, and uranium. The X and Yattributes represent the grid column and row
respectively of each observation's measurement. After opening the dataset in
the PCP, hide the X, Y, and ObsNo axes.
a. Inspect the distribution of observations about the Uranium axis. How many
sub-populations (peaks) are seen?
b. Adjust the sliders on the Uranium axis to isolate the upper sub-population.
Within that newly isolated sub-population, how many distinct sub-popula-
tions are visible looking at the either the Magnetics or Potassium axes? Are
the visible sub-populations of the Potassium axis the same as those of the
Magnetics axis? Explain your answer.
c. Adjust the sliders on the Magnetics axis to isolate the lower sub-population.
What is the relationship between Magnetics and Potassium?
d. Make a dataset of the isolated sub-population, giving it a name of
“LowMagnetics”. Be sure that the dataset contains all attributes including
ObsNo, X, and Y. Save the newly created derived set to disk to use in a later
exercise.
To summarize, as the parallel plot was introduced, it was described as the
most powerful of the VisMiner visualizations and, as a result, the most complex
in its use. In the preceding examples, it was used to locate patterns between
