TABLE 4. — Clusters and morphological gaps of the combining P4 and UPM coefficient of harmonics datasets. Principal component scores (resulting from EFA and PCA) retained for statistical procedures, clustering algorithms used to determine the optimal number of groups under gap statistic methods (Kmeans, Ward, and centroid indices), Shapiro-Wilk test values for each cluster (α = 0.05), and tolerance regions (γ = 0.05; β = 0.1). Note that the overlapping is calculated using the upper limit of Cluster A and the lower limit of Cluster B on their probability distributions along the canonical axis resulting from the LDA. Note that there are not overlapping between groups, providing statistical evidence for the existence of a morphological gap (Fig. 10A).Morphological gaps dataP4 + UPMPC scoresPC1-17K-means index (N clusters)–0.93 (2)Ward index (N clusters)–0.84 (2)Centroid index (N clusters)–1.23 (2)Cluster A (W / p-value)0.90 / 0.17Cluster B (W / p-value)0.92 / 0.31Upper limit (Cluster A)–2.84Lower limit (Cluster B)2.49