Dear all,
I am struggling with analysing a dataset aimed at explaining invasion success of non-native species. At a country level, I need to relate invasion success (binomial: 0 for failed invasions, 1 for success) to socio-economic variables, taking into account
- Phylogenetic relatedness among introduced species: including a phylogenetic tree
- Country as a random effect
- The fact that data are zero-inflated (most introductions fail).
Any suggestions for R packages that can handle a binomial response variable, phylogenetic trees, random effects and zero-inflation?
Thanks in advance,
Diederik
Dear Diederik,
I am not sure if there is a package that can run this specific analysis, but you could try to use one of these functions to answer your question:
you could use the function 'phyloglm()' to fit a phylogenetic logistic model from the the package phyloglm (http://cran.r-project.org/web/packages/phylolm/index.html).
You also need the 'read.tree()' function from the package ape to read phylogenetic trees into R (http://cran.r-project.org/web/packages/ape/index.html)
The function 'pgls()' from the package caper can also deal with phylogenetic GLS models, but not binomial data. And you can not include random effects in none of these functions.
A third approach would be the 'gls()' function from the package nlme ("This function fits a linear model using generalized least squares. The errors are allowed to be correlated and/or have unequal variances"). You can then use a correlaction matrix calculated with 'corPagel()' from ape to take phylogeny into account.
Further, a nice source of information about packages in R dealing with phylogenetic analysis can be found at this cran task view: http://cran.r-project.org/web/views/Phylogenetics.html
I hope it can help you to find a good approach.
Cheers
You should include the answers you got from the R list serve...
Hi all,
Many thanks for your answers, which allow me to explore my dataset. As suggested above, I report that on the r-sig-phylo mailing list, the MCMCglmm package was recommended to me. Another alternative would be to use stan or bugs for writing your own sampling statements & priors.
Thanks again,
Diederik
Dier Diederik Strubbe ,
I would like to ask you how you managed to solve this problem? I am currently facing a similar one, my data are zero-inflated and binomial, glmm with ZI and betabinomial family solved my issue in the preliminary analysis but I need to include the phylogeny as well.
I am a bit afraid of the MCMCglmm package as it is bayesian and I don't have any experience with such statistics.
Best
Barbora
Hi Barbora,
If you don't want to go Bayesian, you might consider glmmTMB package, which can deal with almost all your problems. There was even a post somewhere on how to include phylogeny into models.
Depending on your question, you could neglect the phylogenetic aspect here (is there a phyl. signal in the data/residuals?).
HTH,
Sergey
I am also looking for a zero-inflated model to deal with data containing zeros. I came across two packages: package "pscl" "(https://rpubs.com/kaz_yos/pscl); and package "glmmTMB" (https://fromthebottomoftheheap.net/2017/05/04/compare-mgcv-with-glmmtmb/). Let's give it a try.
Hello all,
thank you Sergey Rosbakh for the suggestion. I was not aware glmmTMB can deal with phylogeny! I am definitely going to give it a try. I spent the whole morning trying to set a model in MCMCglmm but no luck...
Thanks
Barbora
@ Barbora: no problem, I am happy to help. One further suggestion would be to run your model in the glmmTMB package and then test for residuals autocorrelation (there is a test for that in the DHARMA package). If there are no patterns related to phylogeny, you are good to go. Further, you might consider accounting for phylogeny by including a random effect genus/family. The latter is not an elegant solution for your problem, but might work.
Cheers,
Sergey
Hi Barbora Lepková , I also run into a similar issue some months ago and - lacking any background in Bayesian stats - i was a bit afraid to try MCMCglmm. However, I've found a custom script for this, created by Dr Ben Bolker and Dr Michael Li on github. I cannot find again the link, but i have an example vignette and some scripts, if you're interested. It takes some time to figure out everything and it's quite buggy, at some point i had to use a "fresh" computer to make the models run...
- Badges
- Science topic
- Similar topics
- Filing