At 3:21 AM +0100 1/31/04, Listar wrote:
>Date: Fri, 30 Jan 2004 12:59:16 +0000
>From: "Gareth.J.Martin" <g.j.martin@xxxxxxxxxxxxx>
>Subject: [OM] Re: Updated ebay prices, ave. value
>
>Or you could quote the standard deviation with the average but that's a
>bit too OT ..... ;)!
E*ay prices will not be well-behaved in the statistical sense, or perhaps in
any sense. The simplest standard "robust statistics" for such things are the
median and the interquartile range. (In this, "robust" is a technical term
that means insensitive to deviations from the assumed probability
distribution.) These are comparable to the mean and standard deviation we all
learned in school, and give about the same numerical results on well-behaved
data from gaussian distributions.
Median: The value at which half the samples are below, half are above.
Interquartile Range: The difference in the first quartile and the third
quartile.
First Quartile: The value at which one quarter of the samples are below, three
quarters are above.
Second Quartile: Also known as the Median.
Third Quartile: The value at which three quarters of the samples are below,
one quarter are above.
Fourth Quartile: The maximum reported value.
These are easily computed (just sort the data and read the values at or around
the 1/4, 1/2, and 3/4 points in the sorted list), and have the advantage of
being quite resistant to wild points in the data.
One can also just report the four quartile values.
Many spreadsheets, such as MS Excel, have a function to compute quartiles.
Joe Gwinn
>A very good point though. However its still a
>great piece of work and an excellent guide which I know I'll be
>referring to for some time. I now know what to expect when looking on
>Evil Bay for OM stuff.
>
>All the best,
>Gareth.
>
> > About average value: Arithmetical ave. value in
> > not so much suitable in statistics, it's danger is
> > that one extreme value can distort the result to nonsense
> > value.
The olympus mailinglist olympus@xxxxxxxxxx
To unsubscribe: mailto:olympus-request@xxxxxxxxxx?subject=unsubscribe
To contact the list admins: mailto:olympusadmins@xxxxxxxxxx?subject="Olympus
List Problem"
|