At each step, the stepwise procedure makes a “backward glance” to see if any predictors at an earlier step already included in the model now should be excluded. The value of a predictor is of course depending on all the other predictors. The program conducts a partial F test based on the partial Sum of Squares (SS) for each of the predictors already included in the model and exclude the predictor with the highest p-value that falls short of the significance level specified by ‘F-value exclude’. If the significance level for exclude is higher than include as above, the model will be conservative, in the sense that it will tend to retain predictors included.