ABSTRACT
The analytical performance of a calibration deteriorates rapidly as the accuracy of the concentration values for the training set samples deteriorates. Improvements in the accuracy of a training set’s concentration values can result in major improvements in the analytical performance of the calibration developed from that training set. Making the right choices is particularly difficult because there is no single set of choices that is appropriate for all applications. The best compromise among cost and effort put into the calibration vs. the resulting analytical performance and robustness must be determined on a case by case basis. Generating the calibration is often the easiest step in the whole process thanks to the widespread availability of powerful, inexpensive computers and capable software. Generally speaking, the more validation samples the better. It is nice to have at least as many samples in validation set as were needed in the training set. Ideally, the validation concentrations should be as accurate as the training concentrations.
