Even taking a mean on a binary property doesn't make too much sense. Sure you can do the arithmetic, but what does it mean? Consider that using 0 and 1 for have and don't have are arbitrary choices. You could just as well use 0 and 100, or -1 and 1, or 1 and 0. Any of these would drastically change the standard deviation (and the mean).
For continuous properties, either there is a matural measurement as a dimensionless ratio (rare) or it is dimensioned, in which case the numeric values will change with the units used, but the unit is always part of the answer and the numeric change is merely apparent. For discrete properties with a range (counts), mean and s.d. are often useful, but often a little wierd. The average family may have 2.4 children, bot no real family does! |
