How can you attempt to control for any extraneous variables within a sample experiment on prayer and healing?