R help - Jan 2005 - mixed effects model:how to include initial conditions

Dear R users,

I am analyzing a dataset on growth of plants in response to several 
factors. I am using a mixed-effects model of the following structure:

model<-lme(growth~block*treatment*factor1*factor2,
random=~1|plot/treatment/initialsize)

I have measured the initial size of the plants (in 2003) and thought it 
might be sensible to include this (random) variation into the random 
effects term of the model.

Is that correct? Or should "initialsize" rather be included as a 
covariate into the fixed effects term, as in:

alternative<-lme(growth~block*initialsize*treatment*factor1*factor2,
random=~1|plot/treatment)

I would very much appreciate any suggestions on how to analyze these 
data correctly.

Best regards
Chris.

Christoph Scherber wrote:
> Dear R users,
> 
> I am analyzing a dataset on growth of plants in response to several 
> factors. I am using a mixed-effects model of the following structure:
> 
> model<-lme(growth~block*treatment*factor1*factor2,
> random=~1|plot/treatment/initialsize)
> 
> I have measured the initial size of the plants (in 2003) and thought it 
> might be sensible to include this (random) variation into the random 
> effects term of the model.
> 
> Is that correct? Or should "initialsize" rather be included as a 
> covariate into the fixed effects term, as in:
> 
> alternative<-lme(growth~block*initialsize*treatment*factor1*factor2,
> random=~1|plot/treatment)
> 
> I would very much appreciate any suggestions on how to analyze these 
> data correctly.
> 
> Best regards
> Chris.
I think you should include it as a covariate but not in the way you have 
written it.  I would include it as a separate term, not in an interaction

alternative<-lme(growth~initialsize+block*treatment*factor1*factor2,
  random=~1|plot/treatment)

I recommend that you look carefully at the number of coefficients that 
you need to estimate in the model as you have specified it and perhaps 
change to an initial model that had more additive effects and fewer 
interactions.

Christoph,

If you take a look at journal articles related to this topic published
elsewhere,
you will find that the most common analysis is:

model<-lme(growth~initialsize+block+treatment+....)

maybe it would be important to check if your model really needs the interactions
you pointed out. I would suggest to try simpler models first....which generally
work OK in analysis of growth of plants, trees, etc

CM
>-- Mensaje Original --
>Date: Fri, 21 Jan 2005 12:30:07 +0100
>From: Christoph Scherber <Christoph.Scherber at uni-jena.de>
>To: r-help at stat.math.ethz.ch
>Subject: [R] mixed effects model:how to include  initial conditions 
>
>
>Dear R users,
>
>I am analyzing a dataset on growth of plants in response to several 
>factors. I am using a mixed-effects model of the following structure:
>
>model<-lme(growth~block*treatment*factor1*factor2,
>random=~1|plot/treatment/initialsize)
>
>I have measured the initial size of the plants (in 2003) and thought it
>might be sensible to include this (random) variation into the random 
>effects term of the model.
>
>Is that correct? Or should "initialsize" rather be included as a 
>covariate into the fixed effects term, as in:
>
>alternative<-lme(growth~block*initialsize*treatment*factor1*factor2,
>random=~1|plot/treatment)
>
>I would very much appreciate any suggestions on how to analyze these 
>data correctly.
>
>Best regards
>Chris.
>
>______________________________________________
>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