Global statistics are still useful. Knowing the average numbers of earthquakes bigger than 7 per year in the world is interesting even if the rate they occur is not the same as the rate they occur on the fault at the bottom of your garden.

A global statistic is the combination of all the individual ones, and represents the aggregate of their properties, which may be different from the individual case to individual case. In the dice example, if I have two six-sided and two eight-sided, my average total roll is 16, but I can't work that out from studying only one die. But if I want to bet on the percentages, knowing the mean total for a roll of all four is very useful.

Does that make sense?