Parametric regression produces R^2 as a measure of how well the model predicts the sample and adjusted R^2 as a measure of how well it models the population. What is the equalvalent for non-parametric regression (e.g. loess function) ? Ralf