User blog comment:BunsenH/Uncertainties/@comment-25404524-20160923025356/@comment-24577221-20160924060616

It's the number of times there were N of coins, or diamonds, or steps out of the 30 dice in each group. In that chart, there were never 0 coins or diamonds or steps out of 30; never only 1 or 2. Exactly once, there were 3 c/s/d in a group of 30 -- 3 coins, as it happens, in the lines 141-150 group. There were no cases of 4 c/s/d. There were 2 cases of 5 c/s/d: 5 steps in the lines 71-80 group, and 5 coins in the lines 171-180 group. And so on to the right side of the curve, where there were 2 cases of 17 c/s/d: 17 steps in the lines 141-150 group, and 17 steps in the lines 171-180 group.

And that shows you why the treatment isn't really valid, by the way -- the 3 coins and 17 steps in the lines 141-150 group, and the 5 coins and 17 steps in the lines 171-180 group, because having a really low and a really high value in the same group tend to go together. Having a really low value for one item permits a really high value for another item... actually, not just permits it, but tends to cause it. We really should be counting only one kind of item in each group.