R help - Jun 2006 - spatial corStruct in lme

Hi,

I'm fitting a relatively simple growth model to some forest plot data. Two 
species of trees were planted in different mixtures in 10 (nearly-adjacent) 
plots and measured on four occasions over 10 years. The model is 
constructed in terms of the diameter increments (per year; DI) in the 3 
intervals, in which DI is modelled as a function of mid-interval D and DSQ. 
The details of the fixed part of the model are not so important here, but 
four pertinent variables are distance-dependent competition indices: 
C_spp1fromspp1, C_12, C_21, C_22.

The random structure is: random = ~ 1 | PLOT/TREE.

More complex random structures aren't required, there's no obvious 
heterogeneity, and the serial correlations (corSymm) are trivial.

I've been trying to fit a spatial correlation structure using the X, Y 
coordinates of each tree. I've obviously missed a point or so here.

It seems to me that such a structure could be fit for all trees (both 
species) on their average growth over the 3 intervals, by interval, or by 
species by interval.

For now, it would be step forward to get any of these working.

Apart from me learning how to do it, the import of this is that, if the 
CI's are doing what I hope they might, there should be little residual 
spatial correlation.

Thanks,

Lyndon.

Hello Lyndon,

a couple of things you might try - for starters, the use of spatial
correlation structures is documented in Pinheiro and Bates (2002)
"Mixed Effects Models in S and S-Plus" from Springer-Verlag.  So, you
should try that.  Secondly, it's difficult to provide useful advice in
the absence of a small example.  What's best is, for example, if you
send some code snippets that generate sample data, or read it from an
existing package, and then showcase what you've tried.  If you can
then explain the difference between what you get and what you want,
it's easier to provide constructive advice.  You might be able to work
with the Wheat2 data in the nlme package.

Cheers,

Andrew

On Wed, Jun 07, 2006 at 09:35:02AM +1000, Lyndon Brooks
wrote:
> Hi,
> 
> I'm fitting a relatively simple growth model to some forest plot data.
Two
> species of trees were planted in different mixtures in 10 (nearly-adjacent)
> plots and measured on four occasions over 10 years. The model is 
> constructed in terms of the diameter increments (per year; DI) in the 3 
> intervals, in which DI is modelled as a function of mid-interval D and DSQ.
> The details of the fixed part of the model are not so important here, but 
> four pertinent variables are distance-dependent competition indices: 
> C_spp1fromspp1, C_12, C_21, C_22.
> 
> The random structure is: random = ~ 1 | PLOT/TREE.
> 
> More complex random structures aren't required, there's no obvious 
> heterogeneity, and the serial correlations (corSymm) are trivial.
> 
> I've been trying to fit a spatial correlation structure using the X, Y 
> coordinates of each tree. I've obviously missed a point or so here.
> 
> It seems to me that such a structure could be fit for all trees (both 
> species) on their average growth over the 3 intervals, by interval, or by 
> species by interval.
> 
> For now, it would be step forward to get any of these working.
> 
> Apart from me learning how to do it, the import of this is that, if the 
> CI's are doing what I hope they might, there should be little residual 
> spatial correlation.
> 
> Thanks,
> 
> Lyndon.
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
-- 
Andrew Robinson  
Department of Mathematics and Statistics            Tel: +61-3-8344-9763
University of Melbourne, VIC 3010 Australia         Fax: +61-3-8344-4599
Email: a.robinson at ms.unimelb.edu.au         http://www.ms.unimelb.edu.au