On the random allocation design for the control of selection bias in sequential experiments