Evaluation of Errors in Paired Analytical Data by a Linear Model

Abstract - A simple linear model for interpreting duplicate analytical data for a range of values is described. The conceptual model can be quantified using a reduced major axis model rather than the inappropriate traditional least squares approach. The reduced major axis parameters (slope, intercept) and associated errors provide a quantitative basis for: (1) defining bias, and (2) comparing results within laboratories, between laboratories and between sampling/analytical methods. Application of the linear model must be preceded by extraction of obvious outliers or small groups of values that have an inordinately large impact on values of the statistical parameters defining the model. In addition, consideration must be directed to the presence of more than one population of sampling error, as in the case of multiple styles of mineralization being sampled.