12.7 PERFORMING PCA WITH k = 4
Figure 12.10 shows the resulting (i) unrotated and (ii) rotated component matrices for extracting three components. Let us examine the rotated matrix first in Figure 12.10b. Note that the component weights less than 0.5 have been suppressed, to enhance interpretability. The first principal component (RC1 for Rotated Component 1) is a combination of Different Items Purchased and Purchase Visits, which are positively correlated with each other, since their component weights have the same sign. Components can contain combinations of predictors that are either positively or negatively correlated with each other. Had exactly one of the component weights been negative, then that would have been an indication that Different Items Purchased and Purchase Visits were negatively correlated. The remaining principal components are “singletons,” containing only a single predictor each.
![No alt text required.](https://static.packt-cdn.com/products/9781119526810/graphics/images/c12f010a.gif)
![No alt text required.](https://static.packt-cdn.com/products/9781119526810/graphics/images/c12f010b.gif)
figure 12.10 (a) component weights with no rotation, from r. (b) component...