on 11/06/2003 2:27 pm, Martyn Sherriff at [email protected] wrote:
> I would be grateful for some advice or comments on this output from kwallis2
> (Stata 8). On the basis of Kruskal-Wallis I would reject the null hypothesis
> of no difference between regions, but the subsequent multiple comparison
> indicate no significant differences between any of the regions.
The overall anova is telling you that the amount of variation between
regions is larger than you would expect, given the amount of variation
within regions. However, when you start comparing all possible pairs of
regions you run into the problem that no two regions may stand out as
distinctly different.
One way to avoid this kind of problem is to use a regression approach. This
needs a bit of thinking about, but may fit your case:
1. Can you define one of the groups as a natural candidate to be the
control, reference or baseline region? If so, then you can use dummy terms
to test for differences between other regions and this one.
2. If several regions differ from the reference region, you may feel it
appropriate to test if they differ from each other, using a post-hoc test.
The other problem I sometimes encounter (and this probably doesn't apply in
your case, but I mention it for completeness) is an apparently oneway design
in which the categories actually vary on more than one dimension. For
instance, I got some data in to analyse in which the four groups turned out
to be control, drug alone, inhibitor alone and drug plus inhibitor. This is
a two way design, with all the interest in the interaction term, but the
investigators had been analysing it as a oneway design.
Ronan M Conroy ([email protected])
Lecturer in Biostatistics
Royal College of Surgeons
Dublin 2, Ireland
+353 1 402 2431 (fax 2764)
--------------------
--------------------------------------------------------------------------------------------------------------------
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom
they are addressed.
If you have received this email in error please notify the
originator of the message. This footer also confirms that this
email message has been scanned for the presence of computer viruses.
Any views expressed in this message are those of the individual
sender, except where the sender specifies and with authority,
states them to be the views of The Royal College Of Surgeons in Ireland.
--------------------------------------------------------------------------------------------------------------------
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
