Two-way tables: biadditive biplots - Understanding Biplots - page 276

obtained with a call to biad.predictivities and are given in Table 6.5. The overall
quality of the biplots in Figure 6.4 is 61.58%, showing that the interactions are less well
represented in two dimensions than are the main effects. The two biplots in Figure 6.5
are shown separately; we could have incorporated both sets of axes in a single biplot,
but it does not help with intelligibility and therefore is not recommended.

− 100

0

100

− 100

− 50

Cra.H(58)

Tem

0

− 80

Fow.H(12)

Edn.L(88)

50

Beg.L(80)

Ran

− 50

350

20

Kin

− 100

Cra.L(62)

− 50

50

Spo Ear.H(15) 300 300

Edn.H(71)

− 50

250 0

− 150 40

T95

250 − 50

− 150

200

− 60

Hob

0

150

20

200

T64

0 0

− 200

− 200

T68

150

Dur − 20

− 100

Hun

Cap

Fun

100

50

− 100 50

− 50

− 40

0

Ear.L(29)

(a)
Figure 6.7 Similar to the biplot in the top panel of Figure 6.6 but with sites main
effects added by calling the function biadbipl with argument add.maineffects =
TRUE . Note that the scales on the biplot axes have changed to reflect the role of the
main effects. The main effects can also be shown on the biplot axes. This can be imple-
mented by setting argument show.maineffects = TRUE when calling biadbipl .All
main effects are shown as solid black circles on the respective biplot axes. In (a) these
black circles lie on top of one another, coinciding with the origin, and are therefore
not explicitly visible. In the biplot in (b) the point of concurrency of the axes has
been translated in order to improve the display of the biplot. Specifying in biadbipl
the argument select.origin = TRUE allows the user to indicate by means of the
mouse the preferred point of concurrency of the axes. The position of the origin can
be marked with a cross by specifying show.origin = TRUE . Notice how the black
circles indicating the respective main effects have moved together with the axes and are
now readily appreciated. Unlike the biplot in the top panel of Figure 6.6, only the row
points and columns as axes are shown; plotting of the column points is suppressed by
specifying plot.col.points = FALSE and col.points.text = FALSE in the call
to biadbipl .