spatial autocorrelation and GLMMs

[Katharina Ullmann]

Hi all,

I have some count data I'm exploring that shows spatial autocorrelation.

Has anyone incorporated spatial correlation as a random effect in a genearlized linear mixed model? If so, what package did you use and can we talk?

Thanks,

Katharina

[Rosemary Hartman]

One method you can use is just use the GPS coordinates of the sites as your random variable. I haven't done it myself, but i've seen it done, eg:
https://vpn.lib.ucdavis.edu/doi/10.1111/j.1366-9516.2006.00254.x/,DanaInfo=onlinelibrary.wiley.com+full

--
Check out our R resources at http://www.noamross.net/davis-r-users-group.html
---
You received this message because you are subscribed to the Google Groups "Davis R Users' Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to davis-rug+...@googlegroups.com.
Visit this group at http://groups.google.com/group/davis-rug.
For more options, visit https://groups.google.com/d/optout.

--

Rosemary Hartman
roseh...@gmail.com
301-538-0136

[Jaime Ashander]

Package nlme has nice ways to specify covariances in the residuals but it's LMM only. 

One of the best resources for GLMMs is http://glmm.wikidot.com/faq and this suggests you can use the nlme structures via MASS::glmmPQL

There are a variety of other tools mentioned in the section "Spatial and temporal correlation models, heteroscedasticity ("R-side" models)" 

 

- Jaime

[katharina]

Thanks Rosie and Jaime for your suggestions.

I also found this (very new) package called SpaMM that was recently described in a paper by Rousset and Ferdy (2014) Testing environmental and genetic effects int he presence of spatial autocorrelation. Ecography.37: 781-790. It can fit GLMMs with both blocked random effects and spatially correlated random effects. It also supports negative binomial distributions, in addition to poisson and binomial.  I've attached the paper for those who are interested. 

All the best,

Katharina

[ad...@berkeley.edu]

Hi Katharina,

Blast from the past!  I am looking for alternatives to nlme for modeling linear mixed models with spatial autocorrelation, and happened upon your post.  Did you end up using spaMM for your analysis? Could you let me know if you found any good tutorials and resources for it?

Thanks very much!

-Adrian

[Sacha Heath]

Hi Adrian,

An approach that would require some experience with the rethinking package on your part, but with interesting abilities for interpretation, etc. is to use the spatial autocorrelation model presented in the package vignettes  (and discussed a bit more in the book), which incorporates your pairwise distance matrix into your model. 

Sacha

--
Check out our R resources at http://d-rug.github.io/

---
You received this message because you are subscribed to the Google Groups "Davis R Users' Group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to davis-rug+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/davis-rug.

For more options, visit https://groups.google.com/d/optout.

--

Sacha Heath
PhD Candidate, Ecology Graduate Group
UC Davis - Department of Environmental Science & Policy
2132 Wickson Hall | One Shields Avenue | Davis, CA 95616

sachaheath.com
@sachaheath
(760) 223-6895
1219/1226 PES Building

--