R help - Mar 2010 - Please Post Planned Contrasts Example in lme {nlme}

Hi I am running some linear and non-linear mixed effect models and would like to
do some planned contrasts (a priori contrasts)

I have looked in the help and in many forums and it seems possible to do so but
don't understand how to write the function and I couldn't find an
example in Pinheiro and Bates.

 lme {nlme} has a  contrasts argument but I can't understand how to code it.

Here is a simple example of my working lme model:
dependent variable: APHIDS: log (number of aphids)
Fixed effects: DAY (time)   and  TREATMENT ( 6 different aphid treatments)
Random effects: POPULATION (each population is counted through time)

lme.fit<-lme(log(APHIDS)~ DAY*TREATMENT, random = ~ DAY| 			POPULATION,data=
aphid.data)

My hypothesis testing focuses on comparing Treatments ~ 1 and 4 , and eventually
2 vs 5, and 3 vs 6.

Q1:  How can I do planned contrasts on these pairs for the main effect of
TREATMENT?

Q2: Can I also do this for the interaction term (DAY*TREATMENT)?

Q3: Would this code work for a non-linear mixed effect model    nlme( )


P.S. Even just posting a personal example or a link to one would be greatly
appreciated.


thanks for the help

these forums are incredibly helpful

Martin Turcotte
mart.turcotte@gmail.com




	[[alternative HTML version deleted]]

Hi Martin,

Martin Turcotte wrote:
> Hi I am running some linear and non-linear mixed effect models and would
like to do some planned contrasts (a priori contrasts)
> 
> I have looked in the help and in many forums and it seems possible to do so
but don't understand how to write the function and I couldn't find an
example in Pinheiro and Bates.
> 
>  lme {nlme} has a  contrasts argument but I can't understand how to
code it.
> 
> Here is a simple example of my working lme model:
> dependent variable: APHIDS: log (number of aphids)
> Fixed effects: DAY (time)   and  TREATMENT ( 6 different aphid treatments)
> Random effects: POPULATION (each population is counted through time)
> 
> lme.fit<-lme(log(APHIDS)~ DAY*TREATMENT, random = ~ DAY| 		
POPULATION,data= aphid.data)
> 
> My hypothesis testing focuses on comparing Treatments ~ 1 and 4 , and
eventually 2 vs 5, and 3 vs 6.
> 
> Q1:  How can I do planned contrasts on these pairs for the main effect of
TREATMENT?
> 
> Q2: Can I also do this for the interaction term (DAY*TREATMENT)?
> 
> Q3: Would this code work for a non-linear mixed effect model    nlme( )
I would have a look at the multcomp package which allows for
defining contrasts in a very convenient way and can deal a.o.
with lme (from nlme) and mer (from lme4) objects.

http://cran.r-project.org/web/packages/multcomp/index.html

The package comes with several vignettes with example uses
(also an lmer one IIRC).

Best,
Tobias

On 3/18/10, Martin Turcotte <mart.turcotte at gmail.com>
wrote:
> Hi I am running some linear and non-linear mixed effect models and would
like to do some planned contrasts (a priori contrasts)
>
>  I have looked in the help and in many forums and it seems possible to do
so but don't understand how to write the function and I couldn't find an
example in Pinheiro and Bates.
>
Did you try John Fox' appendix on mixed effects models [1]?
Liviu

[1] http://socserv.socsci.mcmaster.ca/jfox/Books/Companion/appendix.html