Databases Reference
In-Depth Information
in the sliders to the right of the plot.When the plot is first viewed, the sliders are set
to the mean value for numeric input attributes and to the modal value for nominal
inputs. For example, in the plot of Figure 6.8, bathrooms is set at 2.36. Therefore,
the model surface of the plot represents homes with 2.36 bathrooms only.
To see the contribution of bathroom count with respect to price, try slowly
dragging the bathrooms slider from the low (left) end at 1.00 to the right. As you
drag, the surface twists and changes shape. Some areas rise and some fall. This
is another indication of interaction between bathrooms, lot size, and square feet.
One way to see the contribution of nominal attributes is to toggle back and
forth between values while noting changes in the plot surface. For example,
toggling den between “N” and “Y” helps to understand the contribution of a den
to a home's price.
Another way to see the contribution of nominal attributes is to select that
attribute in the “Category” drop-down.
Select propertyType in the category drop-down.
The resulting plot is in Figure 6.9. It shows that for almost all combinations,
manufactured homes contribute the least to a home's value, while single family
homes generally contribute the most. Remind yourself again that this is holding
all other inputs constant.
The reader should hereby be warned that any attempts to assess the contribu-
tions of input attributes using two- or three-dimensional plots representing one or
Figure 6.9
Regression Surface by Property Type
