Re: Median: Oasis and Fast Sorting Algorithm

[Leonard Mada <discoleo gmx net>]

Well, the OASIS formula has a big problem with situations like the 
following:
data set is: 1,1,1. So the list contains 4 values of 1. So, the median 
is the middle value, BUT there is really just one value repeated 3 times.

Or consider the following list: 1, 2, 2, 2, 3, 4. So the calculation of 
the median does take just 2 of the 3 values of 2, which is a little bit 
ambiguous. Which one is the middle value? Are 2 two's more middle than 
the 3rd two?

Just my thoughts.

Leonard


Andreas J. Guelzow wrote:
> On Sat, 2007-10-02 at 00:38 +0200, Leonard Mada wrote:
>
>   
> > 2. The *OASIS open formula document* puts the sorting of the array as a 
> > prerequisite to defining the median. This is misleading and should 
> > therefore be replaced with a more algorithm neutral definition. 
> > Unfortunately, the subscription to OASIS is too expensive for me, 
> > therefore I hope that other persons that do have access to the 
> > development board point this out.
> >     
>
> You are clearly misreading the document. The Open Formula document does
> not prescribe algorithms but only describes the return value of the
> Median function. So the value ought to be the middle value (or the
> average of the two middle values) if the data were sorted. No
> implementaion would be required to in fact sort the data!
>
> (...)
>
>   
> > Although this definition is somehow more complex, I believe it is more 
> > accurate and more algorithm neutral.
> >     
>
> How can it be _more_ "accurate"?  The OpenFormula description is 100%
> accurate.
>
> I think it is imperative that the least complex accurate definition is
> used in the OpenFormula document. It is up to implementation to chose
> the preferred algorithm. (The ideal algorithm may vary depending on the
> type of data they typically encounter.)
>
> Andreas
>
>
>
>