2-way Friedman
Evie
1-way Friedman's with no problem.
Thanks
Evie
I want to do Friedman's test (because my data is skewed) rather than
ANOVA. I have repeated-measures data but also
have 2 groups (treatment, control). In SPSS the Friedman's test only
allows you to enter the repeated-measures data but doesn't allow you
to use the groups. Have you any suggestions?
Thanks for your help.
Evie
Martin P. Holt
You could maybe try the SPSSX-L list:
http://www.listserv.uga.edu/cgi-bin/wa?SUBED1=spssx-l&A=1
As well as being very knowledgeable about SPSS, they have a number of good
statisticians.
It is a very active list....
Best Wishes,
Martin Holt
Evie
> >> Thanks- Hide quoted text -
>
> - Show quoted text -
Bruce Weaver
Some skewness is not necessarily fatal. How large are your samples?
Is the direction and degree of skewness similar in all groups or
cells? Here is a quote from Dave Howell's book, "Statistical Methods
for Psychology" (1997 edition, p. 321).
"In general, if the populations can be assumed to be symmetric, or at
least similar in shape (e.g., all negatively skewed), and if the
largest [sample] variance is no more than four times the smallest, the
analysis of variance is most likely to be valid. It is important to
note, however, that heterogeneity of variance and unequal sample sizes
do not mix. If you have reason to anticipate unequal variances, make
every effort to keep your sample sizes as equal as possible."
Granted, Howell is talking specifically about between-groups ANOVA
here, not repeated measures. But AFAIK, the point about all
populations having similar shapes should still be valid.
Finally, if you really do want or need a rank based test, there are
two alternatives I can think of. One is described here:
www.angelfire.com/wv/bwhomedir/notes/alternative_to_anova.txt
Conover (author of "Practical Nonparametric Statistics") suggests that
when sample sizes are large enough, one can simply carry out the usual
parametric test on rank-transformed data.
SPSS does not have built in, single-click procedures for either of
those approaches, AFAIK.
--
Bruce Weaver
bwe...@lakeheadu.ca
www.angelfire.com/wv/bwhomedir
"When all else fails, RTFM."
Bruce Weaver
> Hi Evie,
>
> You could maybe try the SPSSX-L list:
>
> http://www.listserv.uga.edu/cgi-bin/wa?SUBED1=spssx-l&A=1
>
> As well as being very knowledgeable about SPSS, they have a number of good
> statisticians.
>
> It is a very active list....
>
> Best Wishes,
>
> Martin Holt
There is also a usenet group for SPSS users: comp.soft-sys.stat.spss.
John Whittington
>There is also a usenet group for SPSS users: comp.soft-sys.stat.spss.
Indeed there is, but I think you'll find that it is bi-directionally
'cross-linked' with the SPSSX-L list, in the same way that the SAS-L list
is cross-linked with comp.soft-sys.sas. In other words, they are probably
two ways of getting at exactly (almost exactly, the linking never seems to
work perfectly!!) the same thing.
Kind Regards,
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: Joh...@mediscience.co.uk
Buckingham MK18 4EL, UK
----------------------------------------------------------------
Bruce Weaver
> At 21:01 17/10/07 +0000, Bruce Weaver wrote:
>
> >There is also a usenet group for SPSS users: comp.soft-sys.stat.spss.
>
> Indeed there is, but I think you'll find that it is bi-directionally
> 'cross-linked' with the SPSSX-L list, in the same way that the SAS-L list
> is cross-linked with comp.soft-sys.sas. In other words, they are probably
> two ways of getting at exactly (almost exactly, the linking never seems to
> work perfectly!!) the same thing.
>
> Kind Regards,
>
> John
Hi John. That may have been so in the past, but it is not the case
any longer. The recent posts for the SPSSX-L mailing list (which can
be seen at the link given below) do not match up with the recent posts
in the usenet group (the 2nd link).
http://www.listserv.uga.edu/cgi-bin/wa?A1=ind0710&L=spssx-l
http://groups.google.com/group/comp.soft-sys.stat.spss/topics
I know that the usenet group sci.stat.edu used to be linked in the
manner you describe to a mailing list; but that link was broken a few
years ago too.
Cheers,
Bruce
Ray Koopman
In the Friedman test, ranks are done over the repeated measures
separately within each subject, making each subject's mean (k+1)/2
and ruling out the possibility of detecting a group main effect.
In a 1-between, 1-within anova with two groups, the test of the
group main effect is just an ordinary independent-groups t-test
on the subjects' mean (or total) scores. Normality is much less
important than homoscedasticity, especially if the group sizes
differ much.
Also, remember that transformations can create or destroy
interactions. Are you expecting a group x time interaction?
Might there be one, even if you're not expecting it?
If there is one, might it be an artifact of the measurement
scale? Part of the transformation lore is that transformations
the homogenize variances often remove unwanted interactions.
Margaret
On reading your messages, I am a little concerned that there may be
some scope for a conceptual misunderstanding. In particular, the
Friedman test is a non-parametric test for related data and involves
only one group variable. It is sometimes referred to as the Friedman
two-way anova test on account of the manner in which the chi-square
test statistic is evaluated (through interpretation of the ranks for
the data in terms of a two-way design where subjects and treatments
are both used to identify individual ranks). In other words, the term
'two-way' refers to the ranks, not to your own experimental design.
There is not a stand-alone Friedman test which can take the place of a
standard two-way mixed ANOVA when considering non-parametric data. The
following link may be of assistance here:
http://itl.nist.gov/div898/software/dataplot/refman1/auxillar/friedman.htm
As you may already know, within SPSS, there is a command 'Split file'
under the menu 'Data', which allows for sub-group analysis, in this
case performing the Friedman test separately for different treatment
groups. However, when we take such an approach we should be sensitive
to the issue of correcting for multiple comparisons (http://
www.bmj.com/cgi/reprint/310/6973/170).
Hope this helps
Best wishes
Margaret
Evie
Thank you very much, your reply was very helpful. I thought that
"Friedman's test" and "2-way Friedman ANOVA" were the same but
couldn't understand why, so thanks for explaining it to me.
Thanks to everyone for their helpful comments.
Evie
On Oct 18, 10:10 am, Margaret <Margaret.MacDoug...@ed.ac.uk> wrote:
> Dear Evie
>
> On reading your messages, I am a little concerned that there may be
> some scope for a conceptual misunderstanding. In particular, the
> Friedman test is a non-parametric test for related data and involves
> only one group variable. It is sometimes referred to as the Friedman
> two-way anova test on account of the manner in which the chi-square
> test statistic is evaluated (through interpretation of the ranks for
> the data in terms of a two-way design where subjects and treatments
> are both used to identify individual ranks). In other words, the term
> 'two-way' refers to the ranks, not to your own experimental design.
> There is not a stand-alone Friedman test which can take the place of a
> standard two-way mixed ANOVA when considering non-parametric data. The
> following link may be of assistance here:http://itl.nist.gov/div898/software/dataplot/refman1/auxillar/friedma...
>
> As you may already know, within SPSS, there is a command 'Split file'
> under the menu 'Data', which allows for sub-group analysis, in this
> case performing the Friedman test separately for different treatment
> groups. However, when we take such an approach we should be sensitive
> to the issue of correcting for multiple comparisons (http://www.bmj.com/cgi/reprint/310/6973/170).
>
> Hope this helps
>
> Best wishes
>
> Margaret
>
> On Oct 17, 12:11 pm, Evie <sos_st...@yahoo.co.uk> wrote:
>
>
>
> > For further clarification:
>
> > I want to do Friedman's test (because my data is skewed) rather than
> > ANOVA. I have repeated-measures data but also
> > have 2 groups (treatment, control). In SPSS the Friedman's test only
> > allows you to enter the repeated-measures data but doesn't allow you
> > to use the groups. Have you any suggestions?
>
> > Thanks for your help.
>
> > Evie
>
> > On Oct 16, 11:30 am, Evie <sos_st...@yahoo.co.uk> wrote:
>
> > > Does anyone know how to do 2-way Friedman's test in SPSS? I can do the
> > > 1-way Friedman's with no problem.
> > > Thanks- Hide quoted text -
>
> > - Show quoted text -- Hide quoted text -