[R] Error message in gamm. Problem with temporal correlation structure

[Simon Wood]

eva,

The problem is that the random effects and correlation structure that 
you have specified don't meet the restrictions required by lme (which 
gamm calls). You can see this by trying

M0 <-lme(H ~ Tillage , random=list(Block=~1), na.action=na.omit, data = 
mydata, correlation = corARMA(form =~ Year|Tillage, p = 1, q = 0))

which will fail because you have multiple observations of Year for each 
level of Tillage, and because the random effect and correlation formulae 
are incompatible.

The multiple years problem could perhaps be solved by
   corARMA(form =~ Year|Tillage/Block, p = 1, q = 0)
if that makes modelling sense?

A couple of possible fixes are then...

M1 <-gamm(H ~ Tillage + s(Year, by =Tillage), 
random=list(Tillage=~1,Block=~1), na.action=na.omit, data = mydata, 
correlation = corARMA(form =~ Year|Tillage/Block, p = 1, q = 0))

or, if you really don't want that r.e. structure, then

M2 <-gamm(H ~ Tillage + s(Year, by =Tillage)+s(Block,bs="re"), 
na.action=na.omit, data = mydata, correlation = corARMA(form =~ 
Year|Tillage/Block, p = 1, q = 0))

[I've assumed that the mapping between this message and the data you 
sent off list is DepVar=H, Treatment=Tillage]

best,
Simon


On 17/02/12 09:35, HERNANDEZ PLAZA, MARIA EVA wrote:
>
>
> HELLO ALL,
>
> I AM GETTING AN ERROR MESSAGE WHEN TRYING TO RUN A GAMM MODEL LIKE THE ONE BELOW.
>
>   I AM USING R VERSION 2.14.1 (2011-12-22) AND MGCV 1.7-12.
>
> M1<-gamm(DepVar ~ Treatment + s(Year, by =Treatment), random=list(Block=~1), na.action=na.omit, data = mydata, correlation = corARMA(form =~ Year|Treatment, p = 1, q = 0))
>
> THIS IS THE ERROR MESSAGE
>
> Error in `*tmp*`[[k]] : attempt to select less than one element
>
> I have 312 observations. I get the error when I introduce the correlation structure in the model. I have the same problem even if I use /corAR1() /or I set bs=”cr”. I also tried to increase the number of iterations using the control option, but I find the same problem.
>
> Is there any other way in which I could include the correlation structure?
>
> I would appreciate any suggestions
>
> thanks!
>
> eva
> 	[[alternative HTML version deleted]]
>
>
>
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Simon Wood, Mathematical Science, University of Bath BA2 7AY UK
+44 (0)1225 386603               http://people.bath.ac.uk/sw283