[OJFord]

No, what you're getting at is actually exactly why the median is preferred to mean for this sort of thing. The mean is 1.something, it wouldn't round to 1, but you can see how that might happen and be undesirable in other cases.

The median selects the most 'middling' result, it's like the mode (the majority) except sorted and middled: if you had ~as many people with 0 as 2 legs (bimodal), then you'd end up with a median of 1.

[sokoloff]

If 50.1% of people had 2 legs and 49.9% of people had 0 legs, the median number of legs would be 2, not 1.

The median number of testicles in humans is 2. The mean is slightly over 1.

[OJFord]

I didn't mean that 0 people had 1 leg, if 45% have each 0 and 2 legs, and the remaining 10% have 1, then the median is 1. I meant it just as a hypothetical adjustment to the real world, instead of almost 0 with 0, almost 0 with 1, almost everyone with 2; what if it were almost 50% with 0, almost 0 with 1, almost 50% with 2.