
False quantification to the n th degree
SS51000 (SS51@NEMOMUS.BITNET)
Wed, 15 Jun 1994 10:59:40 CDT
Here is an example that shows that the usual rule against showing more
significant digits than your measurement allows need not always apply to
measures of central tendency. Many statisticiansincluding me, using
the term a bit looselywill agree that the median of 1, 2, 3, and 4
is neither 2 nor 3 but the exact midpoint2.500... Note that the mean
also is 2.5. Am I to assume that some of you think that these
calculations are incorrect, or that the median and mean do not exist?
Admittedly, caution is needed: if we know that 1, 2, 3, and 4 here are
ages recorded according to our custom of rounding down (I am "29" even
when I am 29.999), we should add .5 to each value before finding the
central tendency. Bob Graber
