Geoscience Reference
In-Depth Information
A. Estuary
1.5
1
0.5
0
50
100
150
200
250
300
350
400
450
500
1
B. Stream
0.5
0
−0.5
50
100
150
200
250
300
350
400
450
500
1
C. WTP
0.5
0
50
100
150
200
250
300
350
400
450
500
Sample Number
Figure 10.14. Discrimination between estuarine, stream, and WTP samples using PLS-DA, with class
membership indicated by symbols plotting above the dotted line.
(A)
Estuary sites, closed triangles.
(B)
Stream sites, open diamonds.
(C)
WTP sites, dark circles.
they have undergone some kind of treatment or another, or whether they are “good” or
“bad.” A classification model is constructed from a “training” data set in which each
sample has a known category. Subsequently, it is tested by classifying samples in a new
data set using an algorithmically determined distance statistic to predict whether new
samples belong to one category or another. Methods commonly used for this purpose in
chemometrics include
k
-nearest neighbor (
k
NN), a supervised classification method based
on PCA called soft independent modeling of class analogy (SIMCA), and both linear
