Comparison of sampling procedures for investigation of chemical homogeneity of solids by spark emission spectrography
Investigations of the homogeneity of solids are based on the statistical comparison of the total dispersion of values measured on different sampling areas with the error of the analytical procedure. In the case of spectrographic and other destructive methods the procedural error can be estimated only approximately, because independent replicate measurements are not possible. A series of methods, practically used for the estimation of the procedural error, are compared with each other. It is shown that the best approximation of the true error of the analytical procedure will be obtained by multiple sparking of the same sample point and consequently by the same approach used with non-destructive procedures.