The p × p matrix of component loadings (i.e., the correlation coefficients between Zc and Y) is calculated as follows (De Winter and Dodou 2016; Howladar and Rahman 2016): {text{K}}_{text{c}} = {text{ W}}.
where D is the dimension of the feature vector; Γ H ( n ) is the covariance matrix of component n for the hypothesis H; and μ H ( n ) is the vector of the mean values of component n for the hypothesis H.
In PCA, each manifest variable is a linear function of principal components, with no separate representation of unique variance {text{Y}} = {text{Z}}_{text{c}} {text{K}}_{c}^{prime }where, Zc is a N*p matrix of standardized component scores and Kc is a p*p matrix of component loading.
Given the foregoing matrix of component correlations, a second-order PC factor analysis was performed on the eight factor scores (Gorsuch, 1983).
The PCA matrix of component loadings (Table 2) shows the correlation between the original morphological measurements and the two principal components.
The composite matrix A is defined as follows: (3) where A s is the distance matrix of component s, s=1,···, S (number of components) and each submatrix R of appropriate size describes the possible distance values between feature points of two components.
