Multidimensional scaling and nonlinear biplots - Understanding Biplots - page 241

0.2

r

g

RGF

0.3

c

SLF

2

5

p

h

3

i

f

q 2.5

m

j

2

d

u

v

4

4.5

1.5

k

n

e

0.4

1

w

t

s

6

SPR

a b

4

0.5

PLF

0.5

Figure 5.29 Nonlinear biplot based on Clark's distance for aircraft data with normal
projection prediction biplot trajectories.

CircularNonLinear.predictions can be called as shown below for aircraft f and
g for returning the calculated predictions of any desired point:

> out <- Nonlinbipl(X = aircraft.data[,2:5]+0.2, ax.type =

"predictive", dist = "Clark", prediction.type = "circle",
colours = "blue", pch.samples = 15, num.points = 50)

>

CircularNonLinear.predictions(out$Z[7,1:2], Biplot.axis =
lapply(out$Z.axes, FUN = function(x)x[,1:2]),
Biplot.markers = lapply(out$Z.axes,
FUN = function(x)x[,3]))$Markers

The return value (i.e. the predictions for aircraft f and g ) of the above call is:

SPR RGF PLF SLF

f 1.6002 4.1932 0.3157 1.1097

g 1.9815 4.8125 0.2744 2.5983