multiple group comparisons with Mann Whitney

multiple group comparisons with Mann Whitney

Quite a few of my results do tend to be non-normal and I am using mann
whitney or kruskall wallis.

However mann whitney only allows me to do pairwise comparisons and kruskall
wallis only gives me a significance as a whole (if you see what I mean) is
there a way to run mann whitney on a range of groups with all possible
pairwise comparisons, or do I need to cut and past the syntax and just
laboriously edit it.

Rob

multiple group comparisons with Mann Whitney

Quote:> Quite a few of my results do tend to be non-normal and I am using mann
> whitney or kruskall wallis.

> However mann whitney only allows me to do pairwise comparisons and kruskall
> wallis only gives me a significance as a whole (if you see what I mean) is
> there a way to run mann whitney on a range of groups with all possible
> pairwise comparisons, or do I need to cut and past the syntax and just
> laboriously edit it.

How large are your sample sizes? If they are large enough for the z-tests
approximations of those tests, then you could simply use the independent
t-test on ranks in place of Mann-Whitney, and one-way ANOVA on ranks (with
the appropriate multiple comparison procedures) in place of
Kruskall-Wallis.  These tests will be almost identical to the z-test
approximations, if the samples are "large".

See Conover's "Practical Nonparametric Statistics" (1999, 3rd Ed) for more
info.

Cheers,
Bruce
--
Bruce Weaver

Homepage:   http://www.angelfire.com/wv/bwhomedir/

multiple group comparisons with Mann Whitney

thanks for that unfortunately my group size is only 12 max

:-(

> > Quite a few of my results do tend to be non-normal and I am using mann
> > whitney or kruskall wallis.

> > However mann whitney only allows me to do pairwise comparisons and
kruskall
> > wallis only gives me a significance as a whole (if you see what I mean)
is
> > there a way to run mann whitney on a range of groups with all possible
> > pairwise comparisons, or do I need to cut and past the syntax and just
> > laboriously edit it.

> How large are your sample sizes? If they are large enough for the z-tests
> approximations of those tests, then you could simply use the independent
> t-test on ranks in place of Mann-Whitney, and one-way ANOVA on ranks (with
> the appropriate multiple comparison procedures) in place of
> Kruskall-Wallis.  These tests will be almost identical to the z-test
> approximations, if the samples are "large".

> See Conover's "Practical Nonparametric Statistics" (1999, 3rd Ed) for more
> info.

> Cheers,
> Bruce
> --
> Bruce Weaver

> Homepage:   http://www.angelfire.com/wv/bwhomedir/

multiple group comparisons with Mann Whitney

On Mon, 14 Apr 2003 20:30:07 +0100, "Rob Lambkin"

> thanks for that unfortunately my group size is only 12 max

> :-(

[ snip, rest]

Samples of 12?
- things look odd to you?
- it seems safer that way?
- you are suspicious about data generation?

Samples of 12 don't lend themselves to strong
conclusions about their shapes.  You have certain
hypotheses and assumptions, which have not been
spelled out.  So, I wonder if you are (a) overly concerned;
or else, (b)  taking insufficient precautions?

Keep in mind this worry, that the assumptions for testing
with  *ranks*  are, in practice, almost as hard to meet
(about similar shapes) as the assumptions for  ANOVA.

Again, what is the reason that you are worried?
Why do you think it will be  enough to transform to ranks?
- You are saying that actual averages don't matter, but
average-ranks  *do*? (and you don't mention outliers....?)

--

http://www.pitt.edu/~wpilib/index.html

I collected data for five neighbourhoods (using five point scale
questions) and have been told that my data should be treated as
non-parametric. A key research question is to determine if the
differences amongst the five neighbourhoods are significant with
respect to the five point scaled questions. Incidently I conducted 575
surveys (with each survey containing 45 closed-ended questions).

Based on the non-parametric structure of the data, I conducted the
Kruskal-Wallis H test (using SPSS 11) to determine if the five
neighbourhoods differed but this test did not clearly tell me which
neighbourhoods differed! Would it be valid for me to conduct a
Mann-Whitney U test amongst the various pairs of neighbourhoods to
determine which differed significantly (seems like a lot of work as
compared to a traditional post-hoc analysis following an ANOVA test?

Anyway can these two tests then be reported as being used?

Thanks and I apologize if this question is too basic for the group but
I am a stressed "qualitative" geography grad student trying to
complete a long thesis period!

Any help would be appreciated!