The usual methods of calculating the average of a series of samples, as used in the foregoing examples, embody the assumption that from each channel to the next the grade of ore changes at a uniform rate or, what amounts to the same thing so far as numerical results are con¬cerned, that each assay represents the value of the ore for an interval extending halfway to the next sample on each side. Although such an assumption usually yields a perfectly satisfactory approximation, it rarely if ever is in strict accordance with the facts and it can lead to serious error if one or a few samples are notably richer than the rest, a condition which is not uncommon in precious-metal ores and not un-known, though less common, in base-metal ores.

Consider a list of channel samples taken along a drift on a gold-bear-ing vein:
$5.25, $4.00, $17.85, $480.10, $49.20, $22.40, $6.00, $10.15, $1.40, $.70
The arithmetical average of these ten samples, including the high sample assaying $480.10 is $59.70. Omitting the high sample, the average of the remaining nine is $12.99. The critical part that the high sample plays in determining the average prompts the question: is it proper to include such a sample at its full value in averaging a series? In general it is not. Neither is it correct, as a rule, to ignore it.

First, of course, one must eliminate the possibility that the high assay is simply a mistake arising through faulty splitting of the sample or through accidental salting in the laboratory. Careful reduction and reassay of the duplicate half of the original sample should settle this question. Recutting the sample is not always a f air check because some orebodies owe their value to scattered high spots; to recut high samples would eliminate these but might be unfair to the orebody because if the low samples had also been recut they might have disclosed high values where the original sampling did not show them. Some earlier authorities recommend discarding high samples altogether on the ground that this affords a factor of safety. The answer to this is that an under-estimate of grade is almost as misleading as an overestimate; in any case a safety factor should not be a hidden one but, where used, should be introduced deliberately and labeled clearly in the estímate.

How to deal with erratic high samples is one of the knottiest problems in ore estimation. Since attempts to establish standard procedures have not proved wholly satisfactory, most authorities dismiss the subject with a statement that choice of a method must depend on judgment and experience. Truer words were never written, but they are cold comfort to those whose experience in such matters is limited. In order to build up a background of judgment it is well to examine some of the conditions which give rise to erratic high assays.