The experimental regression design : an inquiry into the feasibility of nonrandom treatment allocation