[R-sig-ME] correctly specifying temporal pseudo replication in a mixed effects model

Oyomoare Peters oyomoare at yahoo.com
Wed Jul 10 19:23:16 CEST 2013


Hello,

I would like to correctly specify that my data is temporally pseudo
replicated in a mixed effects model in R using the lme function of the nlme
package. I have done this following the example in Crawley's R text book,
but I need some confirmation that the fixed and random effects are specified
correctly. I have checked previous postings on similar questions and I see
conflicting information. Hence, I decided to ask specifically on this forum.
Also, included in the model is a covariate controlling for spatial
structure, and I would like to confirm that it should be a fixed effect. 

The structure of my data: 
The response variable is turnover in species composition (Tsc) between two
censuses in 26 transects in a tropical forest. Transects are grouped based
on their logging status into four levels (unlogged, lightly logged,
moderately logged, and heavily logged). There are three time intervals for
which turnover was computed (0, 7, 14). Transects of the same logging status
occur in the same location, so Latitude is included as a covariate to
control for logging status. My interest is in the effect of logging status
and time on turnover.

The model I constructed following Crawley's R text is as follows:

m1<-lme(Tsc~Logging.status*Time.interval+Latitude,
random=~Time.interval|Transect, data=mydata)

My questions:
Is it right for Time.interval to be in both the fixed and random effects?
How do I correctly specify that each transect has repeated measures for
three time intervals? 
Should Latitude be a random effect since I am not really interested in it
per say, but only need it to control for spatial structure?

Thanks for your anticipated help. 

Oyomoare



More information about the R-sig-mixed-models mailing list