Discussion - Estimate and Incorporation of Metallurgical Data in a Mineralization Model Technical Papers, MINING ENGINEERING, VOl. 36, No.8 March 1984, pp. 270-275

C. Hertzler The difference of means of milling characteristics by rock and ore type is tested in a statistical manner in the paper. The null hypothesis being tested is that the sample means of two different populations are equal. It is assumed that the null hypothesis is true when performing the test. Because of this, acceptance of this hypothesis leads to the conclusion that the sample means of the two different populations are equal (or there is no significant difference between the sample means), not that the samples come from the same population, as the author asserts. Rejection of the null hypothesis leads to the conclusion that the sample means of the two different populations are not equal. A P value for this test is reported in Table 2 and is incorrectly identified as the probability that the two samples came from different populations. Actually, what was recorded was 100(1 - p) where p is the usual p-value. The usual p-value measures, on a scale of zero to one, the extent to which the data agree with the null hypothesis. It is the probability of observing a value of the test statistic (Z score) that is at least as contradictory to the null hypothesis as the one computed from the sample data (McClave and Dietrich II, 1982). The author makes no mention of the fact that were these tests to be conducted together, the chances of determining that the null hypothesis is not true when it is really true are increased considerably (Snedecor and Cochran, 1980). This is a multiple comparisons problem and should be dealt with accordingly. The usual methods involve adjusting the level of the individual tests. Table 2 reports an incorrect 100(1 - p) value in the comparison of concentrate grades for other rock types with .30 to .69 % Cu vs other rock types with .70 + % Cu; 86 % should be 98.6 %. Although the article can be criticized for poor data presentation and discussion of that data, the point the author makes regarding the use of these statistics in the ore reserve estimation process is well taken. It does not appear that inaccuracies in the statistical interpretations have hindered that usefulness. References McClave, J.T., and Dietrich, II, F.H., 1982, "Statistics," Dellen Publishing Co., San Francisco and Santa Clara, CA, 766 p. Snedecor, G.W., and Cochran, W.G., 1980, "Statistical Methods," The Iowa State University Press, Ames, IA, 561 p. Reply by V. Miller After reading Ms. Hertzler's comments, it appears that the differences noted were basically semantics. If the paper is ever presented at a math symposium, some changes will be needed. When deciding the best way to present the data to mining engineers and geologists, I decided that presenting the probability that the two sample groupings came from different populations (reverse to the Null hypothesis) was easier and would require less verbage to explain. Hence, the 100(1 - p) difference which was noted. Ideally, I would have preferred to give the recovery and concentrate grades, then comment which ones had significant differences. Unfortunately, Kennecott Corp. did not grant permission to publish the ore body's actual recoveries and concentrate grades. The conclusion is that there was, as expected, a relationship between the ore's geologic properties and the metallurgical properties, and by determining this relationship, the overall mineral model was greatly improved.