If there’s one thing I’ve learned through many “non-scientific” studies is to make sure you have enough plants in your different groups to reasonably prove “statistical likelihood.” Basic genetic differences from plant to plant can radically alter the size, quality and harvest of a given plant, assuming the same treatment. With small plant counts (e.g. 1-2), it’s very easy to see these differences, yet they have zero bearing to your test. We usually work with 4+ plants in each different group and often, that’s not even enough to prove statistical likelihood. “Real” studies typically perform averages across dozens, hundreds or thousands of plants.