|
Models |
|
|
For any entry-by-tester
two-way data, GGEbiplot provides 4 options of data transformation,
4 option of data centering, 4 options of data scaling, and 3
types of singular value partition, resulting in 192 biplots of
different shapes. Each biplot has different interpretations and can be
useful depending on research purposes. |
|
|
Data
Transformation |
|
|
|
|
|
0. No
transformation |
|
|
|
|
1.
Natural log |
|
|
|
|
2.
Log10 |
|
|
|
|
3.
Square root |
|
|
|
Scaled
(divided) By |
|
Each value in the
to-way table is divided by some properties of the testers so that the data
is somehow "standardized.". |
|
|
|
0. No
scaling |
|
|
|
|
1.
Tester Std Deviation |
All tested as treated
equally important in evaluating the entries. |
|
|
|
2.
Tester Std Error |
Any heterogeneity among
testers is removed by this scaling. This is the desired option if replicated
data are available. |
|
|
|
3.
Tester LSD5% |
Similar to above but the
tabulated data is more meaningful. |
|
|
Centered By |
|
The means (main
effects) of the entries and/or testers are removed from the biplot. |
|
|
|
0. No
centering |
|
|
|
|
1.
Global-centered (E+G+GE) |
|
|
|
|
2.
Tester-centered (G+GE) |
This results in the
recommended GGE biplots for mega-environment analysis, genotype evaluation,
and test environment evaluation. |
|
|
|
3.
Double-centered (GE) |
This results in the GE
biplot, which contains only genotype by environment interaction. |
|
|
Singular Value Partition |
|
Different options are
equally valid in visualizing the entry-tester interactions, but they lead to
different shapes of the biplot, which have different interpretations.
|
|
|
|
1.
Entry-metric (f=1) |
This biplot is most
appropriate for entry evaluation. |
|
|
|
2.
Tester-metric (f=0) |
This biplot is most
appropriate for tester evaluation. |
|
|
|
3.
Symmetrical (f=0.5) |
This option has been the
most used but it is least useful. It is not ideal for either entry
evaluation or tester evaluation. |
