Why is it more accurate to compare the average height gain of the control and experimental groups (instead of comparing individual plants)?