4.3. Statistical Analyses

Analyses of variance (ANOVA) of the characteristics M100, Prot, and mineral contents in whole grains were performed using the following mathematical model:

where Y_ij is the observed value of the experimental unit containing cultivar i in block j; µ is the general average; c_i is the effect of cultivar i; b_j is the effect of block j; and ε_ij is the experimental error. ANOVA in split-plot schemes for mineral contents in the different grain components was performed using the following model:

where Y_ijk is the observed value of the experimental unit containing cultivar i, seed tissue j, and block k; µ is the overall mean; c_i is the effect of cultivar i; t_j is the effect of seed tissue j; b_k is the effect of block k; ε_ik is the error associated with the plots (error a); ct_ij is the effect of the interaction between cultivar i and tissue j; and ε_ijk is the error associated with the subplots (error b). The means of the treatments were grouped using the Scott–Knott test [37] at 5% probability level.

The relationship between characteristics was verified by simple Pearson linear correlation analysis using the correlation network approach. The phenotypic divergence between cultivars of Andean and Mesoamerican origin was verified by conducting principal component analysis (PCA) and Ward’s hierarchical clustering [38], based on the standardized mean Euclidean distance. The ideal number of groups formed in the dendrograms was established using the Mojena’s method [39]. Statistical analyses were performed with R software (https://www.r-project.org/) using the “Exp.Des” [40], “ggplot2” [41], “factoextra” [42], “pheatmap” [43], and “qgraph” [44] packages.