It seems to me that interval arithmetic is a somewhat different and simpler problem than simulated annealing for optimization. There are many semi-quantitative fields -- e.g., social science -- where data on measurement error is sparse. The thoughtful use of interval arithmetic can give a sense of the precision -- or lack thereof -- of a result. For example, if a population census is thought to have a net undercount of 3 to 5%, what impact will that have on a calculated birth or death rate? How many digits or decimal places to retain?
Sorry to mention the competition, but Mathematica has a convenient Interval [min,max] function that handles this problem quite nicely.
My proposed solution in Mathcad answered the original question, but it is very limited -- it works for multiplication, but not for division, addition, or subtraction.
PS: Also sorry to 'exhume' and old post, but it was there.
TKB
