Use the recommendation of Vladislav Susoy... also try to introduce λ in the regression analysis (PGLSλ, sensu Revell 2010), where phylogeny is incorporated as a variance- covariance matrix with λ in the error term of the regression equation. The error term is then decomposed into a component that represents the phylogeny and the remaining error term. In this approach, when λ is forced to be equal to 0, it is equivalent to OLS regression (i.e. a species level analysis in which the phylogeny is not considered). On the other hand, when λ is forced to be equal to 1, the results are similar to those obtained with phylogenetically independent contrasts (PIC). However, when the estimation of λ is between 0 and 1, neither OLS nor PIC methods are suitable given that they underestimate or overestimate the influence of phylogeny (see Revell 2010; Hernandez et al., 2013). 

Article Phylogenetic signal and linear regression on species data

Article Using phylogenetic information and the comparative method to...

You need to split your question differing taxonomic differences of plant species, from plant communities' differences. To reveal phylogenetic lines (for plant taxa) you need to use any of cladistics ideologies and tools, for evolution of the communities you may use Brooks' or Ronquist's ideology and tools. Just try search here, in the ResearchGate network

thanks Alexander. 

to be clear, I just want to compare species, not community.  for example, i would like to compare growth rate of evergreen tree vs. deciduous tree in a large forest plot with many tree species.

i am wondering, are there a easy way to sovle the problem? I use R for data analysis often, thus I think there are maybe some codes ready for this. even though i know a little about it,  too much Phylogenetic theory and software analysis is still intimadating  to me.

can anyone suggest more  specific on this?

Jingming

To correct for phylogenetic non-independence you could use the threshold model (Felsenstein, 2012), which is implemented in the R package phytools, or Bayesian MCMC sampling as implemented in BayesTraits (Pagel and Meade, 2013). For these tests you would need a phylogenetic tree for all taxa of interest (you can also run simulations using several trees, if you want to account for phylogenetic uncertainty) and a table of trait variables for those taxa.

great ! thanks!

can you tell me what' fuction i can use in phytools? phyANOVA? or in BayesTraits?

I have got phylotree and taxa data ready, and I calculated the PICs,  but PICs could not be compared directly. I tried phy.ANOVA, however, I still not sure it is proper as my trait is highly skewed, not following normal distribution.

wait for your response while try to find the paper you recommended.

Jingming

the function for the threshold model is threshBayes. You can find more information on how to use it here:

http://www.phytools.org/mpma/Exercise_16.1/

http://blog.phytools.org/2012/08/bayesian-mcmc-threshold-model.html

And you can find more about this method in the original paper:

http://www.ncbi.nlm.nih.gov/pubmed/22218305

for my question,  I really think the phytools package is problematic.

still, it lead me to geiger, I will try aov.phylo

....

thanks all.

the paper (Felsenstein, 2012) is really clear, however, this make me doubt the reliable of Brownian evolution assumptioin. the author rejected result of this model, and suggest use other models, such as OU model or set limit to traits value, I don't know how to make it happen for my data. still need to figure out

phylogenetic least squares (PGLS) is very possible solution to me, I tried corBrownian model in APE and it is ok, however,  when I try OU model by corMartins, it don't work, ......still need to figure out

I would say, some of the packege guild is not so friendly useful for beginners.

i.e., CAPER seems right, however, I try the sample data, it is ok. when I try comparative.data() to my own data, it told me there are double genus names in it so that the function can not work. still I need to figure out why.

would anyone suggest anything? 

Jingming

PIC (phylogenetic independent contrasts) can do that. There are R package for this.

The only method I know by myself is phylogenetic contrasts, where you calculate residual values from your data by using the assumed topology of the phylogramm to correct variances for phylogenetic dependency. These residual values may then be fed into ANOVA, but you loose some degrees of freedom, of course.

Use the recommendation of Vladislav Susoy... also try to introduce λ in the regression analysis (PGLSλ, sensu Revell 2010), where phylogeny is incorporated as a variance- covariance matrix with λ in the error term of the regression equation. The error term is then decomposed into a component that represents the phylogeny and the remaining error term. In this approach, when λ is forced to be equal to 0, it is equivalent to OLS regression (i.e. a species level analysis in which the phylogeny is not considered). On the other hand, when λ is forced to be equal to 1, the results are similar to those obtained with phylogenetically independent contrasts (PIC). However, when the estimation of λ is between 0 and 1, neither OLS nor PIC methods are suitable given that they underestimate or overestimate the influence of phylogeny (see Revell 2010; Hernandez et al., 2013). 

Article Phylogenetic signal and linear regression on species data

Article Using phylogenetic information and the comparative method to...

thanks Cristián E Hernández.

At this moment, may i ask a specific small question: how to prune a large phylogenetic tree into a simpler one with only orders or families on it, just for save space in ploting the tree with R?

There are some options:

http://blog.phytools.org/2011/03/prune-tree-to-list-of-taxa.html

http://etetoolkit.org/docs/2.3/tutorial/tutorial_trees.html#pruning-trees

Also, you can do this easily in Mesquite software using the “Prune clade” Tool… all is graphically oriented. (http://mesquiteproject.org/mesquite2.7/mesquite_folder/docs/mesquite/trees.html#tools)

Thanks.

athough two of you recommend the blog.phytools.org for reference, it is not accessible from my IP, maybe for my country's reason.  i have to find ways to overcome this...

besides, i can drop.tip() to remove all species, but i can not drop all nodes to keep all orders in the tree...

Jingming, you can collapse or extract parts of a tree easily using Dendroscope:

http://dendroscope.org

thanks Vladislav! I will try this.

may I ask another specific question:  as my phylo tree is too big to manipulate for PGLS, I want to use family phylo mean to regress one trait vs. others. I realize that the family mean is not arithmatic mean, or the family pics is not the average of species pics in the family.  some paper(Zhang et al, 2010, GEB) mentioned that could be solved by "analysis of traits" in PHYLOCOM or "pic3" in  R package picante, the latter is removed and the former is too much for me.  would you recommend other ways or literature?

thanks in advance.

Jingming