Premium
The Precision of “Rapid” Rock Analysis And the Homogeneity of New USGS Standard Rock Samples
Author(s) -
CLEMENCY Ch.V.,
BORDEN D. M.
Publication year - 1978
Publication title -
geostandards newsletter
Language(s) - English
Resource type - Journals
eISSN - 1751-908X
pISSN - 0150-5505
DOI - 10.1111/j.1751-908x.1978.tb00225.x
Subject(s) - homogeneity (statistics) , homogeneous , mineralogy , standard deviation , sample (material) , analytical chemistry (journal) , replicate , relative standard deviation , oxide , statistics , geology , dissolution , mathematics , chemistry , materials science , environmental chemistry , detection limit , chromatography , metallurgy , combinatorics
Additional analytical data for the thirteen oxides usually determined in a rock analysis by wet methods and estimates and, conclusions from the analysis of variance are presented for the eight new USGS rock standards STM‐1, RGM‐1, QLO‐1, SCo‐1, MAG‐1, SDC‐1, BHVO‐1 and SGR‐1. The “single‐solution” method of sample dissolution, and the “rapid methods” described by Shapiro were used. Three bottles of each standard were analyzed in duplicate, providing six replicate analyses of each sample. Relative standard deviations ranged from 0. 00 to 202 %, with only 12 of the possible 104 sample‐oxide combinations exceeding 10 %. Most deviations exceeding 10 % occur with oxides at very low concentrations or with the determi‐nation, of H 2 O + and CO 2 , all of which often yield poor precision. These data indicate that the rapid wet chemical methods, except possibly for H 2 O + and CO 2 , or when very low concentrations are present, can be considered reliable. Of the 104 F‐ratios calculated during the analysis of variance for sample‐oxide combinations, only eight equal or exceed the tabled value of 9. 55 for F 0.95 (d. f. 2, 3). These data strongly support the conclusion that the contents of the bottles may be considered homogeneous for most oxides determined. When our data for an oxide in the several samples are plotted versus similar data in the USGS Prof. Paper 840 by several methods, the plots generally show good agreement for most sample‐oxide combinations, but approximately thirty of the 104 show substantial differences, with no obvious correlation between the various methods used or the nature or magnitude of the differences.