I am currently examining the scaling of bone length and circumference in different groups of theropod dinosaurs. At the moment I have split the groups and carried out separate major-axis regressions for each group. I've also used dummy codes for each clade and tested whether these improve regression models using step-wise regression. However, it has been suggested that I do an analysis to directly correct the variables for phylogenetic relatedness using something like independent contrasts. My understanding is that variables corrected in this way have to be regressed through the origin (to account for the fact that branches can be rotated around the nodes). As the intercept is of interest when studying allometry, I was wondering if there are any other ways to phylogenetically correct the data which would still produce an interpretable intercept?
Have you tried using pgls? There are a number of packages in R are relatively easy to use ('ape', 'geiger') you basically end up adding phylogenetic relatedness as the correlation in either a generalised least squares equation or even linear mixed models (using code from the 'nlme' or lme4' packages). I believe that these methods allow you to keep the intercept (see Rohlf, F. J. 2001. Evolution 55, 2143-2160).
Same advice, the contrasts by their very definition are constrained to a null intercept. Try PGSL with a linear function.
I think that what you get when using the phylogenetically independent contrsats is the slope relating two variables along your phylogenetic hypothesis. For a given data set the intercept (assuming the slope determined using the contrasts method) could be determined afterwards. As far as I have seen some packages have the limitation thet the phylogenetic hypothesis has to be fully resolved (maybe 'ape')...
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology