Fools rush in within such territory, but I would try
a standard repeated measures anova on such data,
despite the measurement scale. Then I would repeat
that on the most nearly Gaussian equivalent of the data,
which could be -invnormal()- applied to the ridits. The
ridits (which have many other names) are calculated
by a function within -egenmore- from SSC. Ideally,
there should be a consistent story. If not, the data
may be too ornery to do much with.
In addition, do plot the data.
Nick
[email protected]
Paul Fenner
> I have been asked to analyse the following experiment The
> outcome score is a
> bacterial concentration scored 0 to 5. The data consists of
> an initial
> score, the score after treatment A and then the score after
> treatment B.
> This is measured on 32 subjects. The whole experiment is then
> repeated on a
> new set of 25 subjects but this time the sequence is initial,
> treatment B,
> treatment A. The null hypothesis is that the removal of bacteria is
> independent of the treatment sequence.
> I have searched the archive but I cannot find a
> non-parametric repeated
> measures anova. As an alternative I can show, using the the
> ranksum test,
> that there there is no significant difference between the
> three measures for
> the two experiments but I am not happy with this as it is
> ignoring the
> correlated nature of the data in each experiment.
> I would be grateful for any advice on handling this data in Stata 9.2.
*
* 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/
