Contents - Index
Under Biplots, click AMMI1 biplot.
Note: The AMMI1 biplot plots main effects of the entries and testers against their IPC1 scores (IPC = Interactive PC). This biplot does not have the two most important properties of a true biplot. Therefore, many interpretations of a biplot do not apply to this AMMI1 biplot.
The two axes of the AMMI1 biplot should be visualized separately. The X-axis represents the mean values of the entries and of the testers, and the unit is the square root of what is in the original data. Thus, from the above AMMI1 biplot, the entries are ranked on their means across testers as follows:
Entry7 > Entry5 > Entry2 > Grand Mean > Entry1 > Entry4 > Entry6 > Entry3,
and the testers are ranked on their means across entries as follows:
Tester5 > Tester6 > Tester8 > Tester2 > Tester1 = Tester7 > Grand Mean > Tester4 > Tester3 > Tester9.
The IPC1 axis represents the interaction scores, thereby specific interactions between the entries and the testers can be visualized. The AMMI1 biplot is effective in summarizing the total variation of the data but it does not show which-won-where, which is one of the important features of GGE biplot analysis (models 1-3).
Note that the units in this biplot is both square root of the original units in the data. This makes the shape of the AMMI1 biplot objective. In conventional AMMI1 biplots, the X-axis is in the units of the original data whereas the Y-axis is in square root of the original data, making the shape of the biplot totally subjective.