Dear Nicole

In a recent paper of mine, I have also dealt with both allometry and
phylogeny, in a different way than Prof. Dean suggests. You might want to
have a look; it might give you some ideas. Here is the link:
http://link.springer.com/article/10.1007%2Fs13127-015-0238-2. Let me know
if you do not have access and I will send you a PDF.

Best wishes,

Anneke van Heteren

2016-02-02 19:26 GMT+11:00 Adams, Dean [EEOBS] <dcad...@iastate.edu>:

> Nicole,
>
>
>
> There is no need to perform multiple residual-based analyses. If you wish
> to obtain shape residuals where both the phylogeny and allometry (size)
> have been taken into account, these are found as residuals from the PGLS
> analysis: shape~size|phylogeny.
>
>
>
> However, a question then is what will these size/phylogeny shape residuals
> be used to investigate?  If the intention is to then evaluate these
> relative to some other factor (say groups), then the correct approach is
> just to perform a factorial PGLS analysis, where: shape~size+factor |
> phylogeny (shape is a function of size and some other factor, given the
> phylogeny). Additionally, if that factor describes groups, you may wish to
> include the size:group interaction term.
>
>
>
> In fact, if that is indeed the case, it is advisable NOT to perform the
> analysis in piecemeal fashion, where residuals from one regression are then
> used in a subsequent linear model to test other effects. The reason is that
> if there is some interaction between model effects (say, between size and
> groups), then the residuals from the first regression are not correctly
> capturing the observed patterns of variation.  This is the multivariate
> equivalent of the ANCOVA problem, and why an ANOVA on residuals from a
> regression is not always the same as performing the ANCOVA analysis. The
> best solution is to simply perform the factorial model, and account for
> size while examining other effects.  For this correct approach, one simply
> requires software that allows one to perform factorial PGLS. Geomorph (and
> as I recall, NTSYS) will allow uesrs to perform factorial PGLS.
>
>
>
> Hope this is helpful.
>
>
>
> Dean
>
>
>
> Dr. Dean C. Adams
>
> Professor
>
> Department of Ecology, Evolution, and Organismal Biology
>
>        Department of Statistics
>
> Iowa State University
>
> www.public.iastate.edu/~dcadams/
>
> phone: 515-294-3834
>
>
>
> *From:* Nicole Dzenowski [mailto:ndzenow...@gmail.com]
> *Sent:* Monday, February 1, 2016 9:18 PM
> *To:* MORPHMET <morphmet@morphometrics.org>
> *Subject:* [MORPHMET] Questions regarding correction for allometry &
> evolutionary allometry
>
>
>
>
> Hi everyone,
>
>
>
> I had some questions regarding allometry correction. I'll be working with
> closely related specimens on a project where I think a moderate chunk of
> the shape variation is due to allometric size differences. Should (or can)
> I correct for both, as in, regress shape on size and take the residuals and
> then use a phylogenetic comparative method on those residuals and the size
> data and then do another multivariate regression and then use the residuals
> from that final regression as my new shape variables?
>
>
>
> Any help or direction is greatly appreciated.
>
>
>
> Thanks!
>
> --
> MORPHMET may be accessed via its webpage at http://www.morphometrics.org
> ---
> You received this message because you are subscribed to the Google Groups
> "MORPHMET" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to morphmet+unsubscr...@morphometrics.org.
>
> --
> MORPHMET may be accessed via its webpage at http://www.morphometrics.org
> ---
> You received this message because you are subscribed to the Google Groups
> "MORPHMET" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to morphmet+unsubscr...@morphometrics.org.
>

-- 
MORPHMET may be accessed via its webpage at http://www.morphometrics.org
--- 
You received this message because you are subscribed to the Google Groups 
"MORPHMET" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to morphmet+unsubscr...@morphometrics.org.

Reply via email to