PREDICTABILITY OF PARTIAL LEAST-SQUARES REGRESSION IN THE DETERMINATION OF COPPER-ALLOYS BY X-RAY-FLUORESCENCE ANALYSIS

The partial least-squares regression method has been applied to the de termination of copper alloys by X-ray fluorescence analysis. Its predi ctive accuracy is satisfactory. The number of standards needed to prep are a partial least-squares regression model is very small. Only 4 sta ndard samples are used to achieve high accuracy when the standard data set includes the highest and lowest element concentrations in the sys tem. When the concentration of an element is beyond the concentration range in the reference standard materials an accurate result can still be obtained by the method. If complex systems are analyzed, such as g eological samples, the combination of the partial least-squares regres sion method with the fundamental parameter method in X-ray fluorescenc e spectrometry can be utilized to make more accurate predictions.