Dear morphometricians,
I would greatly appreciate your advice on analyses of geometric
morphometric data in relationship to geographic distribution and ecology.
I’ve just started giving a look at the literature. These are my first
impression and questions:
1. When relationships between size/shape and latitude/longitude (or
ecological variables) are investigated, two main methods seem to be
used: regression (multivariate multiple for shape) and partial least
square (only for shape). The general aim seems to be the study of
clinal variation and of the influence of ecological variables on
morphology. In this context, I would tend to see latitude/longitude or
ecological variables as independent variables and size or shape as the
dependent ones. However, often a PLS analysis is performed which does
not make a distinction between independent and dependent variables and
looks for patterns of covariation between the two blocks of symmetric
variables. Why is this statistical model used in a context where the
relationship between variables does not seem to be symmetric?
2. A closely related but more general question. If I use a regression
model, I assume that the independent variable is measured without error
(or with an error negligible compared to the dependent variable). What
shall I do if I want to use a regression model (because I am trying to
predict Y using X) but the error in the independent variable is not
negligible?
3. Getting back to the analysis of geographic data, is it appropriate to
use latitude and longitude as variables in a regression or PLS model
even if those variables are angles used to locate a point on a (more or
less) spherical object? The space of the statistical models (standard
parametric regression or PLS) is Euclidean while latitude and longitude
are coordinates of point on a curved space. If I am correct, the use of
latitude and longitude will introduce an error which can be very large
when the analysis concerns geographic variation on a large scale (across
a continent, for instance). Did I get it all wrong?
4. If what I said in the previous points makes any sense, is there a way
for testing this error? May I do something similar (correlation between
matrices of distances) to what is done for testing the tangent space
approximation to the shape space in geometric morphometrics?
5. Last one. Some advice about software. Is there any program which
computes matrices of geographic distances pairwise among all specimens
in a dataset and does this by using latitude and longitude as variables
but measuring the exact distance across the surface of the earth (i.e.,
taking the curvature into account)?
I thank you very much, in advance, for your help, and I apologize for
asking questions which are probably trivial and only reflect my scant
knowledge of the subject.
All the best
Andrea
Dr. Andrea Cardini
Hull York Medical School
The University of York, Heslington, York YO10 5DD, UK
&
The University of Hull,Cottingham Road, Hull HU6 7RX, UK
tel. 01904 321752
fax 01904 321696
E-mail: [EMAIL PROTECTED]
http://www.york.ac.uk/res/fme/people/andrea.htm
------------------------------------------------------------
Sei stanco di girare a vuoto?
Con il nuovo motore di ricerca Interfree trovi di tutto.
Vieni a trovarci: http://search.interfree.it/
Lo Staff di Interfree
------------------------------------------------------------
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
