Statistische vergelijking van twee chemische analyse-methoden met behulp van de gepaarde t-test en regressie-analyse

Abstract

The comparison of two analytical methods is a common task in analytical practice. This report describes the paired t-test and various forms of regression techniques which are commonly applied to test the biases of one analytical method relative to another. The use of the paired t-test and two forms of regression techniques are illustrated by six examples. A LOTUS 1-2-3 macro was developed to carry out the various calculations. The paired t-test is the method of choice if the concentrations are clustered. The simple least squares regression technique and the Deming regression technique are the method of choice if the samples are fairly uniformly distributed over the concentration range. The Deming regression technique is preferred if the standarddeviation at a particular level is high or increases with concentration.