if an exponentiated logistic regression coefficient, i.e. exp(b), is greater than one, when its associated x-variable increases, predicted odds will decrease, holding other x-variables constant.