Are you talking about the degrees of freedom associated with the predictor parameters? If you're doing an actual regression then those should just be 1 for all of them. If you have df greater than one my guess is you have some categorical variables as predictors as well. Is that the case?