Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: stddev -> %

Re: stddev -> %

From: Jan Gelbrich <j_gelbrich_at_westfalen-blatt.de>
Date: Mon, 15 Mar 2004 09:32:02 +0100
Message-ID: <c33ps3$23ld80$1@ID-93924.news.uni-berlin.de>


"Cory Brooks" <coryb_at_nc.rr.com> schrieb im Newsbeitrag news:4t55c.85976$%d3.5480840_at_twister.southeast.rr.com...
> Is there a function in Oracle that will convert variance into a percent
> likelihood? Specficially, I have both an average (A) and a standard
> deviation(SD) for a set of measurements. What I would like to do is
express
> a single NEW measurement (M) as being within x% likelihood.
>
> I believe that |M-A| / SD will give me how many SDs the measurement is
> from the average.

I am not a very specialist for statistics, but AFAIK stddev expresses *one* values for the variance of *all* measurements,
with a reliability of 68% at Sigma 1 (under Gauss distribution curve), and 90% at Sigma 2.
So at those sigmas, the relibility value never changes. And the smaller stddev is, the better are your measurements i.t.o. precision.

Do You rather mean the aberration of a single measurment from the average ? This is *not* its sttdev !

>The function that I'm trying to find would convert that
> into the % of measurements that are that close or closer to the average.

% of what ? There is no conversion from stddev to any meaningful %, and Sigma1 is always 68%.

> Because I would like to do this with some rather small sets of
measurements
> (but treat it as though it came from a large number of measurements), I
want
> to use a calculated value rather than the imprecise value obtained by
> comparing with the acutal values,

What values are more precise in Your case than the actual measurements ? Are You taking into account some a priori values to be met ? In other words: what are You trying to achieve ?

>e.g., if there were 8 values, then new one
> would be placed to within a range of 12.5%. I'd like to derive a more
> precise value depend where the new value lies between the closest existing
> measurements.
>
> Is there such a function or good way to calculate this % value?
>
> TIA
> Cory
>

just curious,
Jan Received on Mon Mar 15 2004 - 02:32:02 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US