Comparison of selected retention models in reversed-phase liquid chromatography

Retention models are usually compared by how well the model equation fits retention data for one solute taken over a range of mobile phase compositions. Even when retention data for multiple solutes are used, the quality of the fit is often judged by the statistical goodness-of-fit alone. This study compared four different RPLC retention models, encompassing three distinct mathematical forms. Each model was fit to the retention data of multiple solutes and the sets of best-fit parameters were examined in terms of the underlying physico-chemical assumptions of the models. Next, for the linear and quadratic models, some of the model parameters were calculated a priori and the rest of the model parameters were then obtained in subsequent fittings. The sets of best-fit parameters obtained in this manner were more consistent with the underlying assumptions of these models than were the sets of parameters obtained entirely through regressions to the experimental data. Thus, the extraction of parameters by fitting a model to the retention data of a single solute may result in unreliable values for those parameters, even in the case of a fit that would be considered good when judged by conventional statistical criteria. That is, although parameters extracted in such a fashion may be suitable for optimization or similar uses, they may not be suitable for determining the appropriateness of the underlying assumptions of retention models.