Databases Reference
In-Depth Information
When looking at scatter plots of the data, shortcomings of the correlation
matrix become apparent. Correlations are one very simple measure of relation-
ships between attributes; yet they hide detail. For example, in the correlation
matrix for the Iris data, you see a relatively strong inverse relationship between
SepalWidth and PetalLength. (Coefficient of correlation is
0.421.) Indicating
that as PetalLength increases, SepalWidth decreases. This is somewhat counter-
intuitive. One would expect that as the size of the flower increases all measures
would increase.
To evaluate this relationship, click on the SepalWidth/PetalLength cell in
the correlation matrix.
Look at the resulting scatter plot (Figure 2.13). Notice the two clusters of plot
points - one below and to the right of the other - which resulted in the inverse
correlation.
Continue your inspection by selecting Variety in the “Category” drop-
down in the options panel above the plot.
Figure 2.13
Sepal Width versus Petal Length
