CHARACTERIZATION OF SOYBEAN CULTIVARS: RAPID HPLC PROFILING BASED ON PROTEIN MARKERS - Soybean and Wheat Crops: Growth, Fertilization, and Yield - page 76

Table 2. Classification of soybean varieties in the training, evaluation, and cross-
validation of the discriminant model proposed.

Size American European African Asian

Training
American 35 32 (91.4%) 0 0 3 (8.6 %)

European 5 0 5 (100%) 0 0

African 1 0 0 1 (100%) 0

Asian 13 1 (7.7%) 1 (7.7%) 0 11 (84.6 %)

Evaluation
American 5 4 (80%) 0 0 1 (20 %)

European 2 1 (50%) 1 (50%) 0 0

Asian 3 0 0 0 3 (100 %)

Cross validation
American

35 28 (80%) 2 (5.7%) 1 (2.8%) 4 (11.4%)

European 5 3 (60%) 1 (20%) 0 1 (20%)

African 1 0 1 (100%) 0 0

Asian 13 3 (23.1%) 3 (23.1%) 0 7 (53.8%)

3.7

Arkansas
Hawaii
Illinois
Indiana
Iowa
Minnessota
Mississippi
Centroids

2.7

1.7

0.7

-0.3

-1.3

-2.3

-3.3

-7 -5 -3 -1 1 3 5

Function 1

Figure 5. Application of discriminant analysis to the area percentages corresponding to 35 different
soybean cultivars from different USA states: Arkansas, Hawaii, Illinois, Indiana, Iowa, Minnesota, and
Mississippi.

The validation of the method was performed by the prediction of the origin of ten
soybean samples observing that eight of them were correctly classified (80% of classification
capability). Moreover, a cross-validation procedure by the treatment of n-1 out of n
observations as training dataset to determine the discrimination rule and the use of that rule
for the classification of the observation left out was conducted with all 14 variables. 36 of 54
soybean varieties were correctly classified observing the best classification for the American
seeds (see Table 2).