I'm studying the relationship between procumbency (protrusion angle of upper incisors) and mechanical advantage of jaw adductor muscles in several ctenomyids (24 species) and octodontids (5 species).
---------------------------------------------------------------------------------------------------
I've measured the phylogenetic signal in all my raw morphological variables (procumbency angle, basilar length, mandibular width, diastema length, out-lever arm, in-lever arms and mechanical advantages of each adductor muscle of the jaw) and i've observed--in general--a weak signal, being the higher of K=0.47.
I specifically need to know if i've to phylogenetically correct my data if I want to test the relationship between procumbency angle and mechanical advantage (in-lever arm/out-lever arm). In this case, this implies the use of Phylogenetic Generalised Least Squares (PGLS). And--if correction is not necessary-- i can test my data using Generalized Linear Models (or Linear Models). But, i'm not really sure which is the minimum value of K starting from which the data have to be corrected.
Hola Alejandra,
I was going to point you in the same direction as Renen did, although I would like to stress two points from that paper that I think are key to what you are asking:
1) It is the residuals of the regression model, not the values of the variables themselves, that need to present phylogenetic signal for PGLS to be needed.
2) In the absence of phylogenetic signal in the residuals, although I agree that you are probably safe doing a PGLS anyways, the technique might perform worse than an ordinary least squares regression.
Beyond that, I don't think there is a minimum value of K above which PGLS methods should be employed. What you can do though is test for the significance of your K value using a randomization procedure, as implemented in function phylosig, package phytools. Another option would be to perform a PGLS estimating the value of lambda simultaneously, which will automatically adjust for the level of phylogenetic signal in your variables (as in function pgls, package caper).
Hope this helps!
Hola Alejandra,
I was going to point you in the same direction as Renen did, although I would like to stress two points from that paper that I think are key to what you are asking:
1) It is the residuals of the regression model, not the values of the variables themselves, that need to present phylogenetic signal for PGLS to be needed.
2) In the absence of phylogenetic signal in the residuals, although I agree that you are probably safe doing a PGLS anyways, the technique might perform worse than an ordinary least squares regression.
Beyond that, I don't think there is a minimum value of K above which PGLS methods should be employed. What you can do though is test for the significance of your K value using a randomization procedure, as implemented in function phylosig, package phytools. Another option would be to perform a PGLS estimating the value of lambda simultaneously, which will automatically adjust for the level of phylogenetic signal in your variables (as in function pgls, package caper).
Hope this helps!
- Badges
- Science method
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics
- Phylogenetic Analysis