A More Detailed Look at Object Scores

A greater insight into the data can be gained by examining the object scores plots labeled by each variable. Ideally, similar objects should form exclusive groups, and these groups should be far from each other.

Figure 1. Object scores labeled with Thread

The plot labeled with Thread shows that the first dimension separates Yes_Thread and No_Thread perfectly. All of the objects with threads have negative object scores, whereas all of the nonthreaded objects have positive scores. Although the two categories do not form compact groups, the perfect differentiation between the categories is generally considered a good result.

Figure 2. Object scores labeled with Head form

The plot labeled with Head form shows that this variable discriminates in both dimensions. The FLAT objects group together in the lower right corner of the plot, whereas the CUP objects group together in the upper right. CONE objects all lie in the upper left. However, these objects are more spread out than the other groups and, thus, are not as homogeneous. Finally, CYLINDER objects cannot be separated from ROUND objects, both of which lie in the lower left corner of the plot.

Figure 3. Object scores labeled with Length in half-inches

The plot labeled with Length in half-inches shows that this variable does not discriminate in the first dimension. Its categories display no grouping when projected onto a horizontal line. However, Length in half-inches does discriminate in the second dimension. The shorter objects correspond to positive scores, and the longer objects correspond to large negative scores.

Figure 4. Object scores labeled with Brass

The plot labeled with Brass shows that this variable has categories that cannot be separated very well in the first or second dimensions. The object scores are widely spread throughout the space. The brass objects cannot be differentiated from the nonbrass objects.