Hi Liz,
Modelisation of ordinal data is quite difficult. You should read Agresti's
book on categorical data to get some important information. If you're a R
(or S-plus) user you can also download Laura Thompson's document to
accompany Agresti's book.
To be simple, a first step consist of using polynomial contrasts to take
into account the ordinal component. You can then analyze your data as
classical response in any model of your choice but the difference will be
that you will look at linear, quadratic or cubic (depending on the number
of levels) tendencies and not at differences between the levels of your
factor as with a classical nominal variable.
Apart from this quite technical point, I have analyze some BB data with a
colleague for a river restoration program. I'm quite doubtful about the
pertinence of such data as this correspond to important transformation of
raw data.
Another way of analyzing this data could be to use beta models as the BB
scale is bounded.
Regards,
Nicolas
Le Thu, 29 Sep 2011 03:44:43 +0200, Liz Pryde <[email protected]> a
écrit:
Hello Ecolog-ers,
I'm analysing my Braun-Blanquet cover data for a group of vegetation
structure variables (in diff habitat types) and am having trouble
finding a
definitive way to do this legitimately with mixed effects models, ANOVAs
and
ordinations.
The issue is that the BB scale is an ordinal scale. Many studies have
simply
taken midpoint values of the scale classes and turned the ordinal data
into
a metric and analysed from there. To me this seems a bit dodgy, given
that
there are really only 9 values that these variables can take, rather than
the 100 values that is implicit in the metric percentage of cover.
I have read Podani's many papers on the issues and potential solutions
for
ordinations but I seem to be finding it more difficult to find literature
with mixed effects models, or any kind of regression-based model.
Has anyone else encountered these problems? Can anyone suggest some
current
literature on the matter?
Thank you kindly,
Liz
--
Nicolas PERU, PhD
33-(0)4 72 43 28 94
06-88-15-23-10
CNRS, UMR 5023 - LEHNA
Université Claude Bernard - Lyon 1
43 Bld du 11 novembre 1918
Rdc Bât Forel
69622 VILLEURBANNE cedex FRANCE
