In ordinary linear regression analyses, we must often transform the dependent variable to satisfy the usual statistical assumptions of normality, variance stability, and linearity. Use of transformations, however, complicates, interpretations of model parameters. In this project, we derive simple interpretations of linear models that include transformations or interaction terms. These derivations do not replace more traditional presentations but complement and enhance them as an aid to understanding results. A key advantage is that they allow the presentation of results on the original, untransformed scale that is familiar to clinicians, rather than on transformed scale that involves logarithms or power transformations. Preliminary results from this project are to appear in Annals of Epidemiology (in press).
