[R] two lme questions

Scott Rifkin scott.rifkin at yale.edu
Thu Mar 18 16:38:40 CET 2004


1) I have the following data situation:

96 plots
12 varieties
2 time points
2 technical treatments

the experiment is arranged as follows:

a single plot has two varieties tested on it. if variety A on plot #1 has 
treatment T1 applied to it, then variety B on plot #1 has treatment T2 
applied to it.  across the whole experiment variety A is exposed to 
treatment T1 the same number of times as treatment T2.

with respect to time points, plots come in 3 kinds. (1) varietyA,
timepoint#1 vs. variety B, timepoint#1 (2) varietyA timepoint #2 vs. 
varietyB timepoint #2, and (3) varietyA timepoint #1 vs. variety A 
timepoint#2

plots and varieties are random samples from a population of plots and 
varieties, so they are random effects.  The technical treatment and 
timepoints are fixed effects.

i am particularly interested in the variance components for variety and 
timepoint within variety, in the estimate for the fixed stage effect and 
in the predictions (BLUP) for variety and stage within variety.

My First Question is about specifying the random part of the lme() 
statement

the fixed part is Measurement~Treatment+Time
the random part, i think, should include random=~1|variety/time or 
equivalently(?) list(variety=~1,Time=~1)

but how do i also specify that plot should be a random effect?  it's not 
nested within variety, nor is variety nested within it.



2) I have fixed effects as above where each only has two kinds (2 
Treatments, 2 Times).  When I use lme and look at the estimates for the 
fixed effects I get output that looks like:

>summary(asdf.lme)
...

Fixed effects: Measurement ~ Treatment+Time
		Value	 Std.Error ...
(Intercept)	xxx	xxx
TreatmentHot	xxx	xxx
TimeEarly	xxx	xxx
  Correlation
...


where my two treatments are Hot and Cold and my two times are Early and 
Late and the xxx are actual numbers.

Why isnt there any line for the Cold treatment and the late time?  is it 
because these necessarily are the opposite of the Hot and Early ones so 
putting them in would be redundant (i.e. Hot+Cold=0, Early+Late=0)?


Thanks much,
Scott Rifkin
scott.rifkin at yale.edu




More information about the R-help mailing list