Evaluation of Accuracy in Analytical Geochemistry

Surendra P. Verma Laboratorio de Energía Solar, IIM-UNAM, Apartado Postal 34, Temixco, Mor. 62580, Mexico


International geochemical reference samples (IGRS) are commonly employed in the evaluation of accuracy of experimental data in analytical geochemistry, particularly in the establishment of new methods, e.g., HPLC for the REE (Cassidy, 1988; Verma, 1991). In some techniques, e.g. XRF, these IGRS are even used to obtain primary calibration curves (Verma et al., 1992). However, a recent compilation (Velasco and Verma, 1995) of rare-earth elements in
well-studied IGRS from U.S.A. shows that the errors in their characterisation range up to 40%. These can be even higher for less-studied IGRS. On the basis of a recent cooperative study (Govindaraju et al., 1994) of two new IGRS (dolerite WS-E and microgabbro PM-S), it is clear that such errors lie between 10% and >100% for most elements. This implies that we need much better estimates on the "true" values of the concentrations of elements in the IGRS if we were to use them in the evaluation of accuracy in analytical geochemistry. Another implication is that although the means (Potts et al., 1992; Govindaraju, 1994; Irmai et al., 1995) (recommended, preferred, consensus, certified, proposed, compiled, usable, or working values), without the corresponding error estimates and a knowledge of the distribution of the sample population from which they were derived, may certainly be useful in selecting proper material for a given study, they have little value in evaluating and establishing techniques, unless the criteria for their nomenclature are clearly stated, such as those proposed recently (Potts and Kane, 1972).

One way to handle this situation would be to detect objectively outliers in the sample population and thus obtain corrected mean and standard deviation on the IGRS, after elimination of such outlying observations. This has been done traditionally in several ways: e.g., method of select-laboratories (Abbey, 1972); use of certain subjective criteria to first eliminate "aberrant" data and then those lying outside 1S, 2S, or 3S (S = sample standard deviation), without any reference to the distribution of the actual sample populations (De la Roche and Govindaraju, 1973; Gladney et al., 1991; Beyers, 1974).

Numerous statistical tests are in fact available (Lister, 1982; Barnett and Lewis, 1983) for outlier detection, which could be applied objectively to the initial sample population. Four of these tests were recommended for routine application in interlaboratory programs involving determination
of trace elements or radionuclides (Dybczynski, 1980). We have also applied recently two of the more powerful statistical tests (sample skewness and kurtosis) for evaluating IGRS (Velasco and Verma, 1995; Verma and Velasco, 1995).

In this work, I present the results of the application of a large number of these statistical tests for outlierdetection and elimination in recently compiled data (Govindaraju et al., 1994) on two new IGRS (WS-E and PM-S) and classify the final mean values as certified, recommended, or provisional, according to the criteria proposed (Potts and Kane, 1992). These results clearly show the utility of such objective statistical criteria in providing "better" estimates on the composition of these IGRS. Using these IGRS, proper statistical tests (Lister, 1982; Barnett and Lewis, 1983; Bevington, 1969; Sutarno and Stegner, 1985; Velasco and Verma, 1995) can then be applied to assess the quality of the new data in analytical geochemistry.


Abbey, S., Geol. Surv. Can. Pap. No. 72-30 (1972).

Barnett, V. & Lewis, T., Outliers in Statistical Data (Wiley, New York, 1983).

Bevington, P.R., Data Reduction and Error Analysis for the Physical Sciences (McGraw, N.Y., 1969).

Beyers, H.P., South Afr. Bur. Stand. Lett. 10/10/40/1 (1974).

Cassidy, R.M., Chem. Geol. 67, 185-195 (1988).

De la Roche, H. & Govindaraju, K., Bull. Soc. Fr. Cer. 100, 49-75 (1973).

Dybczynski, R., Anal. Chim. Acta 117, 53-70 (1980).

Gladney, E.S., Jones, E.A., Nickell, E.J. & Roelandts, I., Geostand. Newslett. 15, 199-396 (1991).

Govindaraju, K., Potts, P.J., Webb, P.C. & Watson, J.S., Geostand. Newslett. 18, 211-300 (1994).

Govindaraju, K., Geostand. Newslett. 18, special issue, 1-158 (1994).

Imai, N., Terashima, S., Itoh, S. & Ando, A., Geochem. J. 29, 91-95 (1995).

Lister, B., Geostand. Newslett. 6, 175-205 (1982).

Potts, P.J. & Kane, J.S., Geostand. Newslett. 16, 333-341 (1992).

Potts, P.J., Tindle, A.G. & Webb, P.C., Geochemical Reference Material Compositions (Whittles, Caithness, 1992).

Sutarno, R. & Steger, H.F., Talanta 32, 439-445 (1985).

Velasco, F. & Verma, S.P., Geostand. Newslett. (submitted, 1995).

Velasco, F. & Verma, S.P., Actas Inageq 1, 187-192 (1995).

Verma, S.P., Geostand. Newslett. 15, 129-134 (1991).

Verma, S.P., Besch, T., Guevara, M. & Schulz-Dobrich, B. Geostand. Newslett. 16, 301-309 (1992).

Verma, S.P. & Velasco, F., Talanta (submitted, 1995).