Hello Agnieszka

You only need to use phylogenetic correction in your analyses if you are aiming to analyze something that would be impacted by evolutionary relationships among the taxa. The phylogenetic correction serves to avoid statistical errors caused for lack of independence in samples.

Let me give you an example to clarify my explanation: I don't know what you are trying to do with your beetles, but suppose your goal is to investigate if their body sizes are influenced by environment (wet, dry, cool, warm, etc.). To do so, you have to use phylogenetic correction, because in this case the body size could be just an effect of ancestry, that is, closely related species can have similar sizes just because they share an evolutionary history. And being so, if you use a correlation analysis "body size vs. environmental conditions" without correcting for ancestry, your answers could be spurious.

In short, if you are studying adaptation you have to use phylogenetic correction, doesn't matter either the number of species are you considering or the kind of test you are using.

Is it more clearly for you now?

I hope I have helped

Best

Hello Agnieszka

You only need to use phylogenetic correction in your analyses if you are aiming to analyze something that would be impacted by evolutionary relationships among the taxa. The phylogenetic correction serves to avoid statistical errors caused for lack of independence in samples.

Let me give you an example to clarify my explanation: I don't know what you are trying to do with your beetles, but suppose your goal is to investigate if their body sizes are influenced by environment (wet, dry, cool, warm, etc.). To do so, you have to use phylogenetic correction, because in this case the body size could be just an effect of ancestry, that is, closely related species can have similar sizes just because they share an evolutionary history. And being so, if you use a correlation analysis "body size vs. environmental conditions" without correcting for ancestry, your answers could be spurious.

In short, if you are studying adaptation you have to use phylogenetic correction, doesn't matter either the number of species are you considering or the kind of test you are using.

Is it more clearly for you now?

I hope I have helped

Best

Yes, you need to apply phylogenetically informed analyses for most biological relevant variables. The reason is that the statistical tests that you mentioned (ANOVA, ANCOVA, regressions) work under the assumption that each observation (in your case, species) is independent of each other. This assumption is violated when closely related species are more likely to be similar than expected by chance (hence, they are non-independent). As mentioned by Dmitry, although not ideal, you can assume equal branch lengths for your analyses. 

I hope this helps. 

Thank you very much gentleman, this will most definitely help us. Those were our assumptions as well, but we wanted to be sure where the general scientific community stands on this subject. Now we just need to face the music and come up with the proper R script for this purpose.

Be aware that a phylogenetic corrections is not alway need it. Depends on the presence of phylogenetic signal in you data. In the case of correlations, you should test for phylogenetic signal in the residuals error, not the data itself, as you can have no signal in the variables but  have it in the residuals and vice versa. This papers explain it :

http://onlinelibrary.wiley.com/doi/10.1111/j.2041-210X.2010.00044.x/abstract

here are some R scripts of how to do this :

http://blog.phytools.org/2012/11/fitting-model-in-phylogenetic.html

The problem is that testing for phylogenetic signal with less than 20 species is problematic.  Here is an exlanation form Garland:

https://stat.ethz.ch/pipermail/r-sig-phylo/2011-July/001471.html

also, see Blomberg et al. (2003). 

In short, you can estimate the signal in your data but you won't know if is significantly different from 0.  

the usual recommendation in this case is to run your test with and without phylogenetic signal and hope you get similar results. 

good luck. 

Thanks Cristian, we will look through those suggestions and try to apply them accordingly. At the moment, we are facing a problem of constructing a tree with the specific length of branches and then testing it against our data (all lineages of the same length (from root to leafs) and branch lengths negatively proportional to the number of species on each branch). Have you heard of an R script implementing such an approach?

Hi Bartosz, not sure what you are asking. Do you want to get an ultrametric tree (as opposed to one with all branch length set to the same, 1 for example)?

compute.brlen from the the ape package will transform your branch lengh. 

mytreeGraphen

You are exactly right - the term "ultrametric tree" slipped my mind so I tried to describe it as best I could. We wanted to use an ultrametric tree in our case (we don't have branch lengths) and we decided that this approach would be better than assigning arbitrary values to all branches. At the moment, we have the topology coded in Newick and we have a plan to meet tomorrow with one of our colleagues more savvy in R to perform those analyses. This will most definitely help us in doing so. Thanks a lot, ones again.

Hi Everyone,

Thanks you a lot for all answers and constructive advices.

Now, I am sure what I should do and I will be able to solve the problem.

P.S My goal is to invesigate if beetles' metabolic rate is influenced by respiratory pattern, how metabolic rate is affected by body composition and so on.

Best,

Agnieszka 

This is my goal as well, what a coincidence ;)

glad i can help. a final tip,   I recommend this book and website for step by step tutorials on how to run PGLS and other phylogenetically corrected tests. Chapter 5 and 6 should be particularly useful. 

http://www.mpcm-evolution.org/ 

The ultrametric tree could be gotten by doing dating, and further doing so will need the fossil data...