METHOD COMPARISON USING REGRESSION WITH UNCERTAINTIES IN BOTH AXES

Method comparison is one of the most important activities within the d omain of method validation. The results of newly developed analytical methods are regressed against those obtained using reference methods f or a series of samples with different concentrations of the analyte of interest. The linear model, based on the bivariate least squares (BLS ) calibration method, and developed taking into account the comparable errors in both axes, should fit a straight line where the intercept i s not significantly different from zero and the slope not significantl y different from one. To check these premises, the joint confidence in terval for the intercept and the slope is usually applied. This articl e compares the results of this method with the joint confidence test f or the intercept and the slope based on the ordinary least squares (OL S) and weighted least squares (WLS) methods, using two real data sets.