Geoscience Reference
In-Depth Information
Table 5.3
Weights and contrasts (with standard deviations) for seven binary patterns related to
gold deposits in Meguma Terrane
Pattern
No.
W
+
S(W
+
) W
S(W
)
C
S(C)
B
S(B)
6.172
0
0.501
1
0.563
0.143
0.829
0.244
1.392
0.283
1.260
0.301
2
0.836
0.210
0.293
0.160
1.129
0.264
1.322
0.267
3
0.367
0.174
0.268
0.173
0.635
0.246
0.288
0.266
4
0.311
0.128
1.448
0.448
1.787
0.466
1.290
0.505
5
0.223
0.306
0.038
0.134
0.261
0.334
0.505
0.343
6
1.423
0.343
0.375
0.259
1.798
0.430
0.652
0.383
7
0.041
0.271
0.010
0.138
0.051
0.304
0.015
0.309
Source: Agterberg et al. (
1994
, Table 4)
Regression coefficients for logistic model (B) and their standard deviations are shown in the last
two columns. First row (pattern No. 0) is for constant term in weighted logistic regression
Table 5.4
Summary of WofE results for Experiments 1-5
Experiment number
1
2
3
4
5
Number of cells (Training)
13,964
14,570
14,177
14,012
14,353
Number of cells (Testing)
42,133
41,528
41,920
42,086
41,727
Observed number of deposits (Training)
23
15
19
24
43
W1 (Geology)
-1.7756
-2.2753
-1.6337
-0.8486
-2.0326
W2 (Geology)
1.3301
0.9772
0.6636
1.6187
1.1406
W1 (PC3)
-1.3151
-1.1538
-0.2909
-1.1480
-0.9356
W1 (PC3)
0.8527
0.4.63
0.3622
1.2633
0.7601
W1 (Magnetics)
-0.5031
-1.2549
-0.0891
-0.6888
-1.5514
W2 (Magnetics)
1.2494
0.4874
0.4297
0.5245
0.2001
Sum of PPs
37.38
18.79
20.04
46.72
63.40
CI
-test probability
0.903
0.676
0.569
0.941
0.886
Adjustment factor
0.6153
0.7983
0.9482
0.5137
0.677
Observed number of deposits (Testing)
67
75
71
66
47
Estimated number of deposits (Testing)
56
43
20
64
66
Smallest PP
0.0000
0.0000
0.0002
0.0002
0.0000
Largest PP for cell with deposit
0.0054
0.0011
0.0032
0.00078
0.0030
Largest PP for cell without deposit
0.0299
0.0053
0.0054
0.0253
0.0162
Source: Agterberg and Bonham-Carter (
2005
, Table 1)
Note:
W1
,
W2
negative and positive weight,
PP
posterior probability,
CI
conditional independence
the use of logits in studies of this type because, basically, the objective is to estimate
probabilities of occurrence of discrete events. Berkson (
1944
) had introduced logits
as an alternative approach to probit analysis in bioassay (see Fig.
12.8
for a
comparison of logits and probits). Probabilities cannot be negative; neither can
they exceed 1, and these conditions can be violated when the linear least squares
method is used for estimating probabilities. In a later study (Agterberg
1974
), the
Abitibi area was made part of a larger study area (Fig.
5.20
) and a distinction was
Search WWH ::
Custom Search