Contents - Index
How does GGEbiplot treat missing cells
This depends on the centering method, i.e., model that is used in generating the biplot.
GGEbiplot first calculates a two-way table, then subject the table to singular value decomposition (SVD), and thereby displays the rank-2 approximation of the table in a biplot.
If the data is replicated, the two-way table is calculated by simply averaging across all replications, regardless of the number of replicates. Therefore, if the number of replicates is different for different treatments (entry-tester combinations), the resulting two-way table can be biased. A solution to this is to use REML to generate a genotype by enivronment table of BLUP (best linear unbiased prodictors) and use this table to generate the biplot. Current GGEbiplot does not have the capability to generate BLUPs. Another software such as SAS may be needed for this purpose.
The number of missing cells reported by GGEbiplot are the number of empty cells in the two-way table, not that in the input table. Before subjecting to SVD, the missing cells are replaced according to the model:
There are other options for filling missing cells under the menu Data.