I’ve been in the mineral exploration and mining industry for most of my professional life, except for a short period during the early 1980’s when, unlike my peers, many who either left the profession or drove taxis to put bread on the table, I worked in the Hi-Fi retail industry. Mineral exploration is probably the most intense application of the scientific method and exploration geologists are generally regarded as the most experienced of all scientific professions in the practice of the scientific method.
So during 2009 I supervised a very expensive drilling operation in the north-east Pilbara region testing a geophysical gravity target. Geophysical modelling indicated that the observed gravity anomaly could be explained by a channel iron deposit of hematite of some 5000 metres by 400 metres by 10 metres 50-60 metres below the surface. As this exploration target was also in the Cretaceous aged artesian Canning Basin, mud drilling was required in order to avoid artesian water spilling out onto the surface. (Artesian water is under pressure).
The drilling operation was completed and failed to intersect the interpreted geophysical target. Some obvious comments:
- There was consensus of the geophysical interpretation; everyone believed that the interpretation was appropriate.
- Drill testing falsified the interpretation: no iron ore was identified.
- The conclusion was that the interpretation was false/wrong and the hypothesis rejected.
- No statistics were computed to verify this result – the absence of iron-ore was blindingly, in your face, obvious.
Note that no statistics were computed to determine whether the result was random or non-random at some arbitrary R level.
In mineral exploration the outcome is always a binary one – yes or no; there are no maybes; in other words it was bloody obvious the test, drilling, falsified the hypothesis, the geophysical interpretation.
However problems occur when non-scientific thinking is applied to the test. In this case the belief that the computer modelling has to be correct is strong, which then leads to the, but specious, conclusion that the drilling was not performed correctly or wrongly, and that the data must therefore be in error. In order to substantiate such a conclusion statistics are often employed to support the conclusion. It leads to the aphorism that anything might be proven by statistics.
So if you need statistics to demonstrate the verity of some or other hypothesis, then that means that the test itself was incapable of testing the hypothesis, because if it was capable, the result would be ‘blindingly obvious’, not requiring a statistical test.
Mind you if the right result arose as a chance event, then the hypothesis is also false, because true hypotheses don’t produce random results because they can’t. If they do, and one then employs statistical gobbledygook to support the conclusion, then the hypothesis has to be false and its proselytiser being quite unscientific in applying the scientific method.