[R-sig-ME] Spatial autocorrelation correction lme4

[Ben Bolker]

    (Note the list is not just me! -- I should have mentioned that when 
I asked you to forward this e-mail to the mailing list ...)

   The CRAN mixed models task view 
https://cran.r-project.org/web/views/MixedModels.html lists the 
following packages:

Spatial models: nlme (with corStruct functions), CARBayesST, sphet, 
spind, spaMM, glmmfields, glmmTMB, inlabru (spatial point processes via 
log-Gaussian Cox processes), brms, LMMsolver, bamlss; see also the 
Spatial and SpatioTemporal CRAN task views.

   These don't all handle GLMMs, but many of them do.  glmmTMB or spaMM 
might be the quickest to adapt.

   There is a vignette about quasi-AIC associated with the bbmle 
package, but it's not very optimistic about the effort it would take to 
get a qAIC out of a glmmPQL fit: 
https://cran.r-project.org/web/packages/bbmle/vignettes/quasi.pdf



On 2023-06-04 4:43 p.m., Roberto Meseguer Rosagro wrote:
> Dear Dr. Bolker,
> 
> As you are one of the authors of the lme4 package and you are so active in Stack Overflow, I'm reaching you for the following reason (any help would be appreciated):
> 
> I'm building a GLMM (family=Gamma) in lme4 package with some spatially autocorrelated data. Is there any correction that can be applied to the model within this package?
> 
> I know the package nlme has this possibility, but it is not suitable for the Gamma distribution. That's why I ran my global model using the function glmmPQL, from the MASS package (is appropriate for GLMM, and you can use the same structures than in nlme package, so I finally could apply the correction). The problem is that, as glmmPQL uses Penalized Quasi-Likelihood, it doesn't compute an AIC nor a BIC criterion to your global model; so then you cannot dredge it to average the models with �AIC < 2, as glmmPQL global model has not AIC!
> 
> That is why I am wondering if there is any workaround to apply a correction for spatially autocorrelated data to my global lme4 model.
> 
> Sorry for the inconvenience and thanks for your attention.
> 
> Best regards.
> 
> Roberto Meseguer
> 
> 
> 	[[alternative HTML version deleted]]
> 
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models