Question: Many studies have been conducted to investigate the ability of cloud seeding to increase rainfall. The detection of even a large effect is difficult because of the large natural variation in rainfall between storms. One study proceeded as follows: Every day in the study season was evaluated on its suitability for seeding. (Seeding has little chance of being effective if there are no clouds.) On the first suitable day, a random mechanism like the flip of a coin was used to determine whether the clouds would be seeded or not. The rainfall in the target area was measured after treatment. On the next suitable day, the opposite action was taken-the clouds were seeded if they had not been seeded on the previous day and were not seeded otherwise. Again the rainfall in the target area was measured. On the next suitable day, the random mechanism was used again, and on the next the opposite action taken. In this way, consecutive pair of suitable days contained both actions.
(a) Explain why randomization is critically important in this experiment.
(b) Explain how blocking was used and under what conditions this blocking might help.
(c) Would it be useful to have available, for each suitable day, a measurement of the rainfall in an area adjacent to the target area? Why? How useful might it be, relative to the blocking?