I have a theory of statistics: if you can double them or halve them and they still work, they are really good statistics.