Most, though not all, previous applications of PCA to analysis of population structure have taken the data to be a matrix where the rows are indexed by populations not by individuals (e.g., [2,15]). We prefer to consider the larger array where the rows are indexed by individuals. Even when we have population labels, it is useful to examine within-population variation, and we also are often able to find outliers in the data. Furthermore, when population labels are available, we can carry out an analysis to check that the labels do correspond to structure that the eigenanalysis has uncovered.
