[R-sig-phylo] size-free morphometrics in phylogenetic framework

tgarland at ucr.edu tgarland at ucr.edu
Mon Apr 6 18:12:05 CEST 2009

6 April 2009

Hi Dan et al.,

	We did the same thing when testing for phylogenetic signal:

Blomberg, S. P., T. Garland, Jr., and A. R. Ives. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717-745.

	Note that the estimates of slopes are indeed unbiased whether you ignore phylogeny (i.e., use conventional statistics and assume a star phylogeny) or use a phylogenetic method (e.g., PGLS).  However, the slope estimates will not be minimum-variance if you use the wrong assumption for a particular data set.  Also, for a given data set, the estimated slopes CAN be fairly different, so it is not really safe to just do conventional statistics for, say, expedience.

	Another issue is what to do about measurement error.  If you independent variable (e.g., body size) contains "measurement error," then you will likely underestimate the slope:

Ives, A. R., P. E. Midford, and T. Garland, Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252-270.

	Therefore, another alternative is to use a slope derived from first (or second) principles.  In other words, if you have a theoretical reason to think the slope should be a particular value, or just a really "good number" based on lots of previous studies with large data sets, then you might just want to impose that slope.  Numbers like 1.00, 2/3, and 3/4 come to mind for various traits.  It may well end up being less "biased" than any empirically derived slope from a particular data set.

	Of course, you can always do things 2-3 ways ...



Theodore Garland, Jr., Ph.D.
Department of Biology
University of California
Riverside, CA 92521
Phone:  (951) 827-3524 = Ted's office (with answering machine)
Phone:  (951) 827-5724 = Ted's lab
Phone:  (951) 827-5903 = Dept. office
Home Phone:  (951) 328-0820
FAX:  (951) 827-4286 = Dept. office
Email:  tgarland at ucr.edu
List of all publications with PDF files:
Associate Director
Network for Experimental Research on Evolution
(A University of California Multicampus Research Project)

  ---- Original message ----

    Date: Mon, 6 Apr 2009 11:53:05 -0400
    From: Dan Rabosky <dlr32 at cornell.edu>
    Subject: [R-sig-phylo] size-free morphometrics in
    phylogenetic framework
    To: r-sig-phylo at r-project.org

    >Howdy folks-
    >I think I'm opening this up for discussion, rather than
    pointing to a
    >specific problem.
    >Suppose I'm conducting an analysis of phenotypic
    evolution and need
    >size-free species measurements. A standard approach might
    be to take
    >species mean values, then regress those values on some
    index of size
    >(e.g., snout-vent length), and treat the residuals as
    >measurements for further analysis. However, I've also
    seen explicit
    >incorporation of phylogenetic info into the estimation of
    >variables. For example, Collar et al (Evolution, online
    early, DOI:
    >10.1111/j.1558-5646.2009.00626.x) used PIC to estimate
    >slopes of traits on size, then forced these slopes on
    regressions of
    >raw species trait values and used the residuals as
    size-free variables.
    >Do we have strong feelings about why/whether the latter
    is the
    >approach that should be used? My intuition, probably
    incorrect, is
    >that the slopes should be relatively robust regardless of
    >phylogeny is used, but not the degrees of freedom.
    >And if this is a serious problem, wouldn't PCA and
    >morphometric approaches also be affected by the
    assumption that
    >species values represent independent data points?
    >Cornell University
    > [[alternative HTML version deleted]]
    >R-sig-phylo mailing list
    >R-sig-phylo at r-project.org

More information about the R-sig-phylo mailing list