>>Just a little technical note,
>>You divide by "n" if your sample represent the entire set of values.
>>If your data is a sample (a subset of the population) then you need to divide by n-1. I think that VFP std dev. is the latter.
>
>No, it uses the population STD, n. You can test it on a few numbers to see. Of course if n grows larger, the difference is negligible.
>
>I'm not sure why it uses population variance, but one reason may be it avoids the problem of a single data point (where n-1 divisor has trouble :)
>
>Also, overlooked in this whole discussion is whether this STD is even applicable here, it relies on the population being approximately Normal in distribution, which it may not actually be. To be technically correct, one should perform a test of normality on samples of data. Statistics is very easy to go astray in and come up with erroneous conclusions.

You are spot on Bruce. My experience, such as it was, was that most data represents a sample rather than the whole population, but I guess in databases the opposite may be more often true.


David