Comments on coupling graphical and regression analyses of ligand effect data

In this paper we show how graphical analysis of ligand effect data for families of ligands should be combined with regression analysis in order to gain a more self-consistent interpretation of regression results. If a steric threshold shows up in the graphical analysis of the data for the trialkyl ligands (AR(3)), then the threshold must show up in regression analysis of the full set of data. Similarly, the dependence of the data for the isosteric Ligands A(p-XC(6)H(4))(3) must have the same dependence on the electronic parameter chi that we find for the regression analysis for the full set of data. Finally, graphical analysis for both AR(3) and A(p-XC(6)H(4))(3) families shows vividly whether steric effects and/or aryl effects belong in the full analysis. Thus, for results of any regression analysis to be acceptable, they must be consistent with the results of graphical analysis.