[R] spatial autcorrelation in glmmPQL

Dan Bebber danbebber at forestecology.co.uk
Tue Aug 31 21:09:31 CEST 2004

I am unable to specify error spatial autocorrelation structure in glmmPQL:

x and y coordinates for sample points at which presence/absence of seedlings
and canopy openness were recorded in different forest stands.

Does seedling density increase with canopy openness?

Initial analysis using glmmPQL with binomial errors and stand as random
> result <- glmmPQL(seedlings ~ canopy, random = ~1|stand, family=binomial,

Semivariogram of residuals from this fit using geoR showed spatial
autocorrelation with range 34.9 m and relative nugget of 75%. Therefore I
tried to create a corStruct object:
> corel <- corSpher(value = c(34.9,0.75), form = ~x+y, nugget=TRUE)
> corel <- Initialize(corel, data = regen)

And specify the structure in a new glmmPQL
> result2 <- glmmPQL(seedlings ~ canopy, random = ~|stand, family =
binomial, correlation = corel, data=regen)

I get the error message:
iteration 1
Error in eval(expr, envir, enclos) : Object "x" not found

(Running R 1.9.1 on Windows ME)

I appear to be implementing corSpher and Initialize incorrectly- any help
greatly appreciated.

Dr. Daniel P. Bebber
Department of Plant Sciences
University of Oxford
South Parks Road
Tel. 01865 275060

More information about the R-help mailing list