The effects of percentile versus fixed criterion schedules on smoking with equal incentive magnitude for initial abstinence.