160 Computing in Geographic Information Systems
the generalised cross validation function (GCV). This method is relatively
robust because the minimisation of GCV directly addresses the predictive
accuracy and is less dependent on the veracity of the underlying statistical
model. TPS provides a measure of spatial accuracy.
8.1.9 Classification Methods
The classification method uses easily accessible soft information (e.g., soil
types, vegetation types, or administrative areas) to divide the region of in-
terest into sub-regions that can be characterised by the moments (i.e., mean,
variance) of the attribute measured at locations within the region of interest
(Burrough and McDonnell [8]). The model for classification is:
ˆz(x
0
) = µ + α
k
+ (8.5)
where
ˆz