Raul Miller wrote:
On 6/27/07, Eldon Eller <[EMAIL PROTECTED]> wrote:
Argument by reductio ad absurdum: Consider the standard deviation of a
data set consisting of a single point. Using N for the denominator gives
a standard deviation of zero, which is correct only if that point is the
entire population.
Sure -- and it IS correct when all members of the population are
the same.
If you have only a single point and it is not the entire population,
then your problem is that you've not collected enough data to
properly represent deviation.
Exactly so. You cannot determine the deviation from a single point that
is not the entire population. Perhaps I am naive, but this is what
indeterminate means to me. It is an answer, and the correct answer in
that circumstance. I recognize Roy Crabtree's point, that 0%0 = 0 in J,
but J is wrong on this point.
Using N-1 gives a standard deviation of 0%0, which is indeterminate,
and is correct if that point is a sample from a larger population.
"Indeterminate" is a non answer, not a correct answer.
And you have the same underlying issue when N is greater
than 1, if your samples happen to have the same values.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
