The toy example revisited Figure 1 shows the artificial data set used previously. It has n = 101 cases, two variables x and y and three concentric groups indicated by the colours of the points. Last time, I added two columns with x and y to expand the data matrix from 101 × 2 to 101 × 4. A PCA on the expanded data matrix gave a scores plot for the first two PCs in which the groups could be separated linearly.