Geoscience Reference
In-Depth Information
Table 12.1 Comparison
of probabilities estimated
for ten cells by means of
original copper potential
determination method
and the jackknife
Location
p (original)
p (jackknife)
SD
32/62
0.45
0.44
0.08
16/58
0.33
0.32
0.04
17/58
0.39
0.38
0.05
18/58
0.01
0.00
0.06
16/59
0.35
0.36
0.04
17/59
0.33
0.33
0.14
18/59
0.37
0.40
0.13
16/60
0.43
0.47
0.04
17/60
0.06
0.00
0.06
18/60
0.06 0.08
Location is given by UTM grid cell numbers; SD¼ standard
deviation of jackknife estimate (Source: Agterberg
1973
, Table 1)
0.03
cross-product of acidic volcanics and mafic intrusives. Its coefficient was
B
(2,6)
¼
10
4
. Dropping the factor 10
4
, the seven new estimates were 0.49583,
0.37170, 0.51601, 0.52055, 0.20898, 0.54010 and 0.41653. The average of these
7 new estimates for 30 control cells is 0.43853, which is less than 0.50954 for all
35 control cells. A result of this type is not surprising because bias should increase
when fewer control cells are used. The randomly selected five control cells can be
regarded as containing volcanogenic massive sulphide deposits that had not been
discovered. The seven pseudo-values obtained for
B
(2,6) are 0.592, 1.337, 0.471,
0.443, 2.313, 0.326 and 1.068, respectively. The jackknife coefficient becomes
0.9356 with variance 0.07198. Similar results were obtained for the other explan-
atory variables.
The eight stepwise regressions resulted in sets of eight estimated values for all
(10
0.50954
10 km) cells in the Abitibi area. Jackknife estimates were obtained for all
individual cells and these were combined for overlapping sets of cells to obtain
estimated numbers of (10
10 km) cells with one or more volcanogenic massive
sulphide deposits per larger (30
unit cells measuring 40 km on a side were used). The result is shown in Fig.
12.2b
.
There is hardly any difference between the patterns of Fig.
12.2a, b
indicating that
both methods of bias reduction (use of control area supposedly without
undiscovered deposits in Fig.
12.2a
and the jackknife in Fig.
12.2b
) are about
equally good. In Chap.
4
, it was discussed in detail that multivariate regression of
occurrences of mineral deposits on rock types and other variables measured for the
surface of bedrock are only valid for undiscovered deposits occurring relatively
close to bedrock surface. Later discovered copper ore was mostly found deeper
down but within the same favorable areas within the Abitibi Volcanic Belt as the
large copper deposits that already were known in the late 1960s.
Table
12.1
shows a comparison of results for ten selected (10
10 km) cells
obtained by the original application of the general linear model of least squares and
the new jackknife results shown in Fig.
12.2a, b
, respectively. These cells constitute
a subsample of the 1,086 cells used for this example. Cell 32/62 contains the
Magusi River deposit that was discovered in 1972 after publication of the original
Search WWH ::
Custom Search