[18964] in s-news-athena
RE: [S] Shapiro-Wilks' test: summary and question (fwd)
daemon@ATHENA.MIT.EDU (Ranjan Maitra)
Wed Aug 18 11:21:59 1999
Date: Wed, 18 Aug 1999 11:16:50 -0400 (EDT)
From: Ranjan Maitra <maitra@math.umbc.edu>
To: S-news <s-news@wubios.wustl.edu>
Message-Id: <Pine.LNX.4.04.9908181115350.6358-100000@param.math.umbc.edu>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
The following is a message I received from Siem Heisterkamp in response
to my earlier post of this morning.
Thanks,
Ranjan
***************************************************************************
Ranjan Maitra, Department of Mathematics and Statistics,
University of Maryland, Baltimore County, Baltimore, MD 21250, USA.
***************************************************************************
\|/ satyamevajayate | tamasomaajyotirgamaya \|/
-*- | -*-
/|\ TRUTH SHALL PREVAIL | FROM DARKNESS TO LIGHT /|\
***************************************************************************
Ph: 410-455-2436 FAX: 410-455-1066 http://www.math.umbc.edu/~maitra
***************************************************************************
---------- Forwarded message ----------
Date: Wed, 18 Aug 1999 15:18:02 +0200
From: Siem Heisterkamp <s.h.heisterkamp@amc.uva.nl>
To: 'Ranjan Maitra' <maitra@math.umbc.edu>
Subject: RE: [S] Shapiro-Wilks' test: summary and question
I wonder on the purpose of testing a null-hypotheis that a sample of
11000 is stemming from a normal ditribution: I bet that it is not. I
also think that, depending on your purposes, it is quite irrelevant.
Because of the large numbers the mean of this sampel will be normally
distributed (with great chance ayway), and performing
an anova will be as good as performing a kruskall-wallis. Simply
because also the Kruskall wallis depends on
normaliy (chi-squarity) of the test statistic.
BUT to make things short: if you really want to perform a non-par test:
al simple non-par tests (wilcoxon, mannwhitney, kruskall wallis,
spearman etc) can be simply mimicked by: ranking the observations
(accoriding to the recipe of the test) and perform a t-test, or anova
(for kruskall wallis) or pearson correlation ON the ranks.
This is simply equivalent (if no ties) and if there are ties, there is
no alternative anyway. I don't think that
e.g. statexact will handle these things either exactly.
Siem Heistarkamp
> -----Original Message-----
> From: Ranjan Maitra [SMTP:maitra@math.umbc.edu]
> Sent: Wednesday, August 18, 1999 3:05 PM
> To: S-news
> Subject: [S] Shapiro-Wilks' test: summary and question
>
>
> Thanks to Tom Gatcliffe and Peter Mandeville who provided functions to
> perform
> the Shapiro-Wilks' test. Tom's routine works for samples under 2000
> while
> Peter's works for samples of size between 5 and 5000.
>
> Francine Cohen suggested use of the commercial package Environmental
> Stats
> for Splus.
>
> Anyway, my sample is of size 11,168. Does anyone know of
> Kruskal-Wallis test
> for such a large sample?
>
> Thanks,
> Ranjan
>
>
>
> **********************************************************************
> *****
> Ranjan Maitra, Department of Mathematics and Statistics,
> University of Maryland, Baltimore County, Baltimore, MD 21250,
> USA.
>
> **********************************************************************
> *****
>
> \|/ satyamevajayate | tamasomaajyotirgamaya
> \|/
> -*- |
> -*-
> /|\ TRUTH SHALL PREVAIL | FROM DARKNESS TO LIGHT
> /|\
>
>
> **********************************************************************
> *****
> Ph: 410-455-2436 FAX: 410-455-1066
> http://www.math.umbc.edu/~maitra
>
> **********************************************************************
> *****
>
> ----------------------------------------------------------------------
> -
> This message was distributed by s-news@wubios.wustl.edu. To
> unsubscribe
> send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
> message: unsubscribe s-news
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
