Hi Thomas

As put by Kim, the usual way to deal with allometry is multivariate
regression. You can then work on the residuals, which stand for the part
of shape variation independent from size variation.

To me the main problem is when you have several samples : regressing the
whole dataset onto centroid size will mix two things : the within and
among samples allometries (i.e. static and evolutionary allometry),
which are not necessarily the same (see Klingenberg 96 in the NATO ASI
book). Theoretically you should first test whether the allometric
directions are similar in your different groups then (and only if the
answer is yes) apply the regression to the whole dataset (the 'pooled
within group' allometry). Unfortunatelly, in my experience, allometric
directions are often different among groups, which makes it very
difficult to get rid of the among groups allometry...

I am not aware of any way of dealing with allometry in this case. There
is a discussion on this in Klingenberg's 96 paper and a very recent
paper re advocate the CPC-Burnaby method

Mccoy MW, Bolker BM, Osenberg CW, et al.
Size correction: comparing morphological traits among populations and
environments OECOLOGIA 148 (4): 547-554 JUL 2006

Hope this helps.

Any suggestions about the multigroup allometry problem would be warmly
welcome!!

Vincent Debat

-- 
Replies will be sent to the list.
For more information visit http://www.morphometrics.org

Reply via email to