R:Fitcontinuous. Comparison Of Models By Aicc Weights (Density Plot Interpretation)

Question

[Alice]

Hello biostars!

Following this tutorial I tried to analyse trait evolution (data, which is in tutorial zip. archive: limbs and anolis trees).

In the part of "Quantifying Uncertainty", as suggested, I considered 10 trees and fitted three models to all of 10 trees.

Models: "Brownian Motion", "Early Burst" and "Ornstein-Uhlenbeck".

The question is whether our model selection is robust to phylogenetic uncertainty.

fitContinuous suggested EB, and the final step in this part of analysis is comparison of models over distribution of my 10 phylogenies by density plot.

Can you please help me in interpretation of results?

My questions are following:

• Is BM model not robust to phylogenetic uncertainty, because of its curve?
• If EB model is robust to phylogenetic uncertainty, how can we explain it by plot curves?
• Another thoughts about plot?
• What does it mean "density"? Is it "roughly "how many trees over one AICc value?

To be honest, I stucked with this plot and do not understand its meaning and interpretation.

[David W]

Not to dissuade you from posting questions here, I'm sure there are plenty of people who can help with statistics and PCMs, but you should also know the R special interest group of phylogenetic methods is very helpful: https://stat.ethz.ch/mailman/listinfo/r-sig-phylo

Answer 1

[David W]

In reverse order:

What does it mean "density"?

The normal statistical meaning - an attempt to estimate an unknown probability density function. it's similar to a histogram of the AICc values for your models across the 10 trees you sampled

Is x model not robust to phylogenetic uncertainty, because of its curve?

In these plots you aren't really looking at whether one or other model is robust, so much as the means by which you've selected one model from a set of possibles (read up on model selection and AIC if you aren't familiar with that process).

From you figure, I'd say that across your ten trees the EB model always has the greatest AICc weight, which is to say its always the best-fitting of the three models you've considered. Thus, your model _selection_ is robust to phylogenetic uncertainty across these 10 trees. It's still possible that the actual _parameters_ of the EB model you've fitted are not robust with regard to phylogeny uncertainty. You should see how the values of the parameters you are most interested in change across your tress