Information Technology Reference
In-Depth Information
Ta b l e 4 . 2 2
Within-class sample predictivities obtained for the first four and last four
samples in CVA biplots (weighted and unweighted) of the copper froth data.
Weighted CVA
S1
S2
S3
S4
S494
S495
S496
S497
Dim_1
0.1043
0.0054
0.0822
0.1243
0.0028
0.0029
0.0086
0.2460
Dim_2
0.1928
0.2369
0.3715
0.2287
0.3854
0.0093
0.0198
0.2891
Dim_3
0.3814
0.5820
0.4569
0.3192
0.6559
0.0305
0.0571
0.3564
Dim_4
0.3848
0.6474
0.4962
0.3718
0.7384
0.0309
0.0710
0.3564
Dim_5
0.3851
0.7232
0.4962
0.3788
0.7543
0.3444
0.1034
0.3636
Dim_6
0.5295
0.7345
0.5900
0.3976
0.8901
0.8934
0.9570
0.4638
Dim_7
0.5934
0.9034
0.6785
0.7548
0.9956
0.9471
0.9729
0.9071
Dim_8
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
Unweighted
I
CVA
S1
S2
S3
S4
S494
S495
S496
S497
Dim_1
0.1057
0.0110
0.0883
0.1257
0.0080
0.0030
0.0104
0.2423
Dim_2
0.1634
0.2711
0.3483
0.2145
0.4116
0.0064
0.0224
0.2798
Dim_3
0.3846
0.5673
0.4793
0.3370
0.6274
0.0298
0.0568
0.3545
Dim_4
0.3848
0.6474
0.4962
0.3718
0.7384
0.0309
0.0710
0.3564
Dim_5
0.4340
0.6795
0.5225
0.3783
0.8935
0.9391
0.8913
0.4285
Dim_6
0.4415
0.7055
0.5313
0.3786
0.9123
0.9416
0.9945
0.4476
Dim_7
0.8523
0.7454
0.8317
0.5096
0.9144
0.9580
0.9968
0.4499
Dim_8
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
Unweighted Cent CVA
S1
S2
S3
S4
S494
S495
S496
S497
Dim_1
0.1058
0.0111
0.0884
0.1257
0.0081
0.0030
0.0104
0.2423
Dim_2
0.1671
0.2557
0.3571
0.2203
0.3924
0.0066
0.0212
0.2811
Dim_3
0.3835
0.5404
0.4769
0.3318
0.5974
0.0302
0.0534
0.3554
Dim_4
0.3848
0.6474
0.4962
0.3718
0.7384
0.0309
0.0710
0.3564
Dim_5
0.4802
0.6994
0.5576
0.4040
0.9242
0.9337
0.9449
0.3947
Dim_6
0.5253
0.7491
0.5879
0.4073
0.9250
0.9892
0.9864
0.4918
Dim_7
0.8728
0.8482
0.8469
0.5822
0.9295
0.9980
0.9985
0.5391
Dim_8
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
time some tools were made according to the
MSAII
methods. As the
MSAIIU
period
followed after the
MSAIIL
period, the shape of
MSAIIU
towards the bottom shows
the most modern trend. Alpha-bags are therefore not only useful for describing clouds
of points, but can give important information on the structure of overlap in the data.
4.11.2 Quantifying overlap
Although the CVA biplot methodology is useful when classifying samples of unknown
origin to a particular class, perhaps its main strength is its potency for describing the
degree as well as the nature of the overlap among the various classes. We illustrate this
