On Sat, 20 Sep 2003 18:36:41 -0400, "Mark R Marino"
<[EMAIL PROTECTED]> wrote:
> Why do you think that the algorithms in these computer packages differ?
> Why aren't they universal?
> Any thoughts or comments?
Robert Dodier wrote it out in detail, but here is the shorter version.
When it comes to distributions that are discrete,
there is not one single definition of 'percentile' that meets
everyone's needs. ('Percentiles' include quartile).
There is not any textbook or organization that sets a
standard that folks feel a need to follow, for a great
many statistical terms, so no single definition has won out.
I think you can find some computer-program definitions
collected in the last three years, by searching the Usenet groups,
sci.stat.* , for centiles or percentiles.
Here is an excerpt from what I posted in August, 2002 -
========
[ in a post from 18 months previous ]
Bruce wrote,
"I don't know the answer to your question, Alan. But while digging
around, I found this interesting collection of comments on quartiles:
http://exploringdata.cqu.edu.au/ticktack.htm "
======= posted by Warren Sarle of SAS, same day.
SAS provides at least six ways to estimate quantiles (none of which
is what Excel does). Five of them are described at
http://sasdocs.ats.ucla.edu/proc/zormulas.htm . A sixth one is
provided in PROC MEANS for very large data sets and is considerably
more complicated than the first five.
========= end of excerpt .
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization."
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
