αρχείο Email: Practice of causal inference with the propensity of being zero or one: assessing the effect of arbitrary cutoffs of propensity scores