Information Technology Reference
In-Depth Information
Hun(62)
50
−
150
Box.H
0
Cap(72)
50
50
−
50
−
100
Box.L
0
Tru.H
−
50
T64(30)
0
0
Tru.L
0
Ear.L
−
50
0
0
Ear.H
−
50
0
50
Hob(78)
100
Cra.L
Beg.L
Fow.H
20
50
0
Edn.H
0
50
−
100
Cra.H
0
50
−
Edn.L
Beg.H
0
−
50
−
100
50
−
150
Spo(44)
Fow.L
100
100
T95(8)
Kin(75)
100
100
50
−
Figure 6.9
Using circular projection to show all interaction predictions pertaining to site
Fow.L
. Note that the main effects, indicated by black solid circles, are all superimposed
at the origin.
It is clear from Table 6.5 that by adding a further dimension to the biplots in
Figures 6.4 - 6.13 several predictivities will be substantially increased. Furthermore, from
the eigenvalues returned by
biadbipl
it follows that the overall quality increases from
61.6% in two dimensions to 72.9% in three dimensions. Therefore it might be advan-
tageous to consider a three-dimensional biplot of the interaction matrix in Table 6.2.
We provide in Figure 6.15 two screenshots of three-dimensional biadditive biplots. The
screenshot in the top panel was made after the following call to
biadbipl
:
biadbipl(t(wheat.data), lam = F, biad.variant = "InteractionMat",
show.origin = FALSE, SigmaHalf = TRUE, dim.biplot = 3,
pch.row.points = 15, row.points.col = c(rep("blue",5),
rep("orange",7)), row.points.size = 1.25, pch.col.points = 17,
plot.col.points = FALSE, column.points.col =
rep(c("green","red"),7), column.points.size = 1.25,
col.text.3d = rep(c("green","red"),7),
axis.col = rep(c("green","red"),7), column.points.text = TRUE)
