The stepwise regression method of selecting predictors for computer assisted multiple regression analysis was compared with forward, backward, and best subsets regression, using 16 data sets. The results indicated the stepwise method was preferred because of its practical nature, when the models chosen by different selection methods were similar in number of variables, variables included, and amount of variance explained. The best subset method worked very well for these data sets, and was recommended for encouraging a non-mechanical selection process by giving many suggested models. The backward method provided a model which explained about as much variance as models chosen by any other method, but this model may have included more variables than necessary. It was not recommended when there is high multicollinearity. The stepwise method was generally adequate except when conditions of multicollinearity, suppression, and sets of variables working jointly do not occur; then it should be used in conjunction with other methods. The forward method was not recommended if the stepwise method is available. It was concluded that the best subsets and backward procedures were the best, and that the stepwise and forward methods should never be used alone in selecting a model. (GDC)