Re: Standard Deviation
- From: Tom Backer Johnsen <backer@xxxxxxxxxxxx>
- Date: Wed, 12 Oct 2005 23:28:59 +0200
Nils Haeck wrote:
Oh yes, it is correct.
Standard Deviation usually uses N - 1. This is based on the fact that N - 1 will have more predictive value than N, if you look at this from a probability perspective.
The "divided by N" value is usually called variance (denoted by greek small letter sigma).
That terminology may perhaps vary. In my world (social sciences) the point of departure is SS (sum of squared differences from the mean). If you divide that value by N or N-1 you get the variance, which is s squared if you divide by N, or sigma squared if you divide by N-1. From that value again, you get the standard deviation by extracting the square root, which is either s or sigma. The first is an expression of the variation of the sample in itself (or regarded as the population), the other is an estimate of the variation of the population from which the sample is supposed to be drawn.
Normally, and with anything but trivial sample sizes, dividing the value by N or N-1 has only theoretical interest. It may have no relevance at all, as when computing a correlation coefficient. In that case, N is only of interest when considering a test of significance for the correlation coefficient, which is a different subject all together.
Tom .
- References:
- Standard Deviation
- From: Ed Dressel
- Re: Standard Deviation
- From: Nils Haeck
- Re: Standard Deviation
- From: Ed Dressel
- Re: Standard Deviation
- From: Nils Haeck
- Standard Deviation
- Prev by Date: Re: Have they left?
- Next by Date: Re: Delphi and VmWare
- Previous by thread: Re: Standard Deviation
- Next by thread: Re: Standard Deviation
- Index(es):
Relevant Pages
|