Ordinary and orthogonal regressions in QSAR/QSPR and chemistry-related studies

A critical examination of different least squares orthogonal methods (OR1-OR4) and of the ordinary least squares (LS) method, which is normally used in QSAR and QSPR studies, and in many scientific and chemistry-related fields, reveals that not always the orthogonal regression methods perform better than LS in the aforementioned fields. Nonetheless the OR methods, whose use in statistics and economics are considered superior in most cases to LS, relying on the minimization of the sum or quadratic orthogonal distances offer an interesting alternative method for obtaining a graphical 'symmetric' representation, which is not rendered by the LS method.