EVALUATING THE PREDICTIVE POWER OF REGRESSION-MODELS

Regression models are routinely developed and used in aquatic sciences for predictive purposes. Although the traditional measures of predict ive power for regression models (r(2), root mean square error) have we ll-defined statistical meanings, they do not necessarily provide an in tuitive measure of the predictive utility of regression equations. It is proposed that an index of predictive power can be developed on the basis of the degree of categorical resolution a regression model can a chieve. This index of resolution power is shown to increase nonlinearl y with the familiar r(2) statistic, even under different distributiona l assumptions. This relationship also shows that the predictive power of models with r(2) less than or equal to 0.65 is low and nearly const ant but increases very rapidly for higher r(2) values, thereby justify ing the search for additional explanatory variables even in models alr eady explaining a large fraction of the variation.