=STDEV(data) should be =SQRT(DEVSQ(data)/(COUNT(data)-1) which is supposed to
be less than DEVSQ(data) by a factor of 1/SQRT(COUNT(data)-1).

If you mis-stated the situation, and STDEV(data) is instead less than
SQRT(DEVSQ(data)/(COUNT(data)-1), then I will guess that you are using Excel
XP (2002) or older, and your data has a small CV (coefficient of variation,
also known as RSD -- relative standard deviation). Older versions of Excel
used a mathematically correct, but numerically unstable algorithm that could
give numerically incorrect results with a small CV or a very large n.

It is had to be more specific without more information about your particular
case.

Jerry

"xyw" wrote:

The StdDev gave number much smaller than the squre root of sum of square of
deviation from the mean. How is it calculated?