Survey Evaluations
Early geochemical exploration surveys were evaluated using simple statistics.
Often one standard deviation above the mean was defined as anomalous. This is
not the best way to evaluate this type of data. This simple approach can mislead
you on both extremes of a survey. A survey in a background area with no
anomalous samples will still yield a mean and a standard deviation which will
identify the highest non anomalous samples as anomalies. Similarly this test in an
anomalous area will only identify the highest anomalous samples correctly
defining many anomalies as background.
Often one standard deviation above the mean was defined as anomalous. This is
not the best way to evaluate this type of data. This simple approach can mislead
you on both extremes of a survey. A survey in a background area with no
anomalous samples will still yield a mean and a standard deviation which will
identify the highest non anomalous samples as anomalies. Similarly this test in an
anomalous area will only identify the highest anomalous samples correctly
defining many anomalies as background.
Probability Plots
A much better approach is the use of another simple but a more effective
evaluation. A probability plot is a graphical representation of data which will
effectively display a gaussian distribution as a straight line. Two straight lines with
differing slopes indicate two gaussian distributions. Below you will find probability
plots for a number of states. Each graph is composed of the many surveys
coming from the listed state starting in 1982 and continuing to the present.
evaluation. A probability plot is a graphical representation of data which will
effectively display a gaussian distribution as a straight line. Two straight lines with
differing slopes indicate two gaussian distributions. Below you will find probability
plots for a number of states. Each graph is composed of the many surveys
coming from the listed state starting in 1982 and continuing to the present.
Method
The vertical axis represents the percentage of the value in relation to all other
values in the set. If the total set has 100 samples the set is sorted from lowest to
highest and the lowest value is assigned 1% and each subsequent sample is
assigned the next percentage with the highest sample being 100%. Sets larger and
smaller than 100 use the appropriate fraction of a percentage. The graph then
uses the sample value on the horizontal axis and the percentage on the vertical
"probability" axis.
values in the set. If the total set has 100 samples the set is sorted from lowest to
highest and the lowest value is assigned 1% and each subsequent sample is
assigned the next percentage with the highest sample being 100%. Sets larger and
smaller than 100 use the appropriate fraction of a percentage. The graph then
uses the sample value on the horizontal axis and the percentage on the vertical
"probability" axis.
Nearly every set of data I have evaluated using this graphing technique has
identified at least a few samples as anomalous. The survey below is one of the
few which indicated only a single background distribution.
identified at least a few samples as anomalous. The survey below is one of the
few which indicated only a single background distribution.
The roll from background to anomaly is always greater in these large state wide
sets. Individual sets from more confined areas will normally make a sharper
change of slope at a smaller range of values. Interestingly all three of these
diverse areas separate background and anomaly at 2.2 ppm I2. Not all areas,
however, have the same background and anomaly distributions. The sets below
demonstrate this.
sets. Individual sets from more confined areas will normally make a sharper
change of slope at a smaller range of values. Interestingly all three of these
diverse areas separate background and anomaly at 2.2 ppm I2. Not all areas,
however, have the same background and anomaly distributions. The sets below
demonstrate this.
All of the data I have run has yielded similar graphs. Below is the first survey I
ran in 1982 for Davis Oil.
ran in 1982 for Davis Oil.
GrayStone Exploration Labs, Inc.
| How To |