R help - Nov 2011 - glht for lme object with significant interaction term

Dear all, 

I'm working on some data from an experiment on the breeding behavior of
birds. In short, I have been measuring how the time spent on performing a
certain task (variable 'mean_on_active') differs over time (variable
'day',
2 levels) across three experimental categories (variable 'treat'; levels
'C', 'R', 'E'). The model shows a significant
interaction between treatment
and day. To have a closer look at this, I would like to do multiple
comparisons for treatment levels within day (i.e. across treatments for each
day in turn). I have browsed the forum for a way of doing this using the
glht function, but haven't found a good solution to my problem. 

Any hints on how to proceed, or on other methods that might be (more)
appropriate? Final model output below, data attached.
http://r.789695.n4.nabble.com/file/n4097865/sub.data.txt sub.data.txt 

Kind regards, 
Andreas


##Final model
m.final<-lme(mean.on.active~treat+day+treat:day,random=1|id,na.action=na.omit)
summary(m.final)

Linear mixed-effects model fit by REML
 Data: NULL 
       AIC      BIC    logLik
  1051.779 1075.757 -517.8896

Random effects:
 Formula: ~1 | id
        (Intercept) Residual
StdDev:    6.276309 5.473167

Fixed effects: mean_on_active ~ treat * day 
                  Value Std.Error DF   t-value p-value
(Intercept)   24.987123  1.792599 75 13.939049  0.0000
treatE        15.661119  2.327329 73  6.729225  0.0000
treatR         1.133678  2.228713 73  0.508670  0.6125
day6-7        -1.068160  1.758899 73 -0.607289  0.5455
treatE:day6-7 -6.650690  2.335567 73 -2.847570  0.0057
treatR:day6-7 -1.495927  2.273071 73 -0.658108  0.5125
 Correlation: 
              (Intr) treatE treatR day6-7 tE:6-7
treatE        -0.736                            
treatR        -0.747  0.572                     
day6-7        -0.482  0.372  0.389              
treatE:day6-7  0.361 -0.522 -0.292 -0.753       
treatR:day6-7  0.370 -0.281 -0.524 -0.774  0.582

Standardized Within-Group Residuals:
       Min         Q1        Med         Q3        Max 
-2.5774244 -0.4585991 -0.1360731  0.3902143  3.9752257 

Number of Observations: 154
Number of Groups: 76 


--
View this message in context:
http://r.789695.n4.nabble.com/glht-for-lme-object-with-significant-interaction-term-tp4097865p4097865.html
Sent from the R help mailing list archive at Nabble.com.

Nov 23, 2011 at 1:27am Andreas wrote:
> I would like to do multiple comparisons for treatment levels within day
> (i.e. across treatments
> for each day in turn)
Andreas,

The following does what you want. To see how/why it works, look at the
vignette to package multcomp, where there is an example. Note that I had to
change "mean.on.active" to "mean_on_active".

##
library(nlme); library(multcomp)
Tdf <- "Your attached data set"
m.final<-lme(mean.on.active ~ treat * day,random=~1|id,na.action=na.omit,
data=Tdf)
Tdf$IntFac <- interaction(Tdf$treat, Tdf$day, drop=T)
m.finalI<-lme(mean.on.active ~ IntFac,random=~1|id,na.action=na.omit,
data=Tdf)
summary(m.final)
summary(m.finalI)
glht(m.finalI, linfct=mcp(IntFac = "Tukey"))

Regards, Mark.

-----
Mark Difford (Ph.D.)
Research Associate
Botany Department
Nelson Mandela Metropolitan University
Port Elizabeth, South Africa
--
View this message in context:
http://r.789695.n4.nabble.com/glht-for-lme-object-with-significant-interaction-term-tp4097865p4099459.html
Sent from the R help mailing list archive at Nabble.com.