R help - Nov 2010 - lme Random Effects and Covariates

1.  I'm attempting to test for Random Effects.  I've grouped the data on
subject (grid) but want to use lme to build the model without subject as a
RE then add it and do anova between the 2 models.  This is the result I get
and it appears it's adding Random Effects.

tmp.dat4 <- groupedData(Trials ~ 1 | grid, data = tmp.dat4)

mod2a <- lme(Trials ~ factor(group_id) + reversal, data = tmp.dat4,
na.action = na.omit, method = "REML")
> summary(mod2a)
Linear mixed-effects model fit by REML
 Data: tmp.dat4 
       AIC      BIC    logLik
  4544.054 4587.718 -2262.027

Random effects:
 Formula: ~factor(group_id) + reversal | grid
 Structure: General positive-definite
                  StdDev    Corr         
(Intercept)       10.505303 (Intr) fc(_)2
factor(group_id)2  9.830679 -0.778       
reversal2          7.106839 -0.275  0.023
Residual           9.995963              

Fixed effects: Trials ~ factor(group_id) + reversal 
                      Value Std.Error  DF   t-value p-value
(Intercept)       23.275874  1.876185 510 12.405960   0e+00
factor(group_id)2 -7.639842  2.151004  72 -3.551757   7e-04
reversal2          7.681495  1.206858 510  6.364869   0e+00
 Correlation: 
                  (Intr) fc(_)2
factor(group_id)2 -0.785       
reversal2         -0.308 -0.015

Standardized Within-Group Residuals:
       Min         Q1        Med         Q3        Max 
-2.6884393 -0.5059063 -0.1892908  0.4944976  2.8477377 

Number of Observations: 585
Number of Groups: 74 

2.  Secondly is this the correct way to add covariates (such as age).

mod2i <- lme(Trials ~ factor(group_id)*factor(reversal) * age, data tmp.dat4,
random = ~ 1 | grid, na.action = na.omit, method = "ML")

-- 
View this message in context:
http://r.789695.n4.nabble.com/lme-Random-Effects-and-Covariates-tp3049181p3049181.html
Sent from the R help mailing list archive at Nabble.com.

Hello, all,

I have a fairly complicated experimental design and would really appreciate some
help on correctly specifying my random effects.  I have a CRD split plot design
with one covariate measured on the whole plot unit, one on the subplot unit and 
2 other covariates measured on the subsamples in the subplot unit.  My design is
actually unbalanced, but if it *were* balanced, I would have 7 REP (replicates)
each of 3 treatments (TRTMNT1 = whole plot unit are forest stands), each split
into 5 TRTMNT2 (subplot units are 25x25m plots), in which 3 to 7 TREES were
measured (subsampling units).  In addition, I have 1 continuous covariate (ELEV
= elevation) measured on the whole plot units (i.e. all measurements in each REP
have the same measure of ELEV), 1 continuous covariate measured on the subplot
unit (DENSITY of all trees in the subplot unit, so all trees in the subplot have
the same measure of density, but each subplot within a whole plot has a
different measure) and 2 continuous covariates (HEIGHT and DBH) measured on the
subsampling units (trees).

I do not understand the syntax of the lme random effects to properly describe my
structure to make sure that the test of the effect of my covariates, ELEV
measured on the whole plot unit and effect of DENSITY measured on the subplot
unit, are correctly specified, with correct df in numerator and denominator,
etc.

What is the syntax for this random effects structure?  How do I indicate that
ELEV is measured on the whole plot level, DENSITY on the subplot and HEIGHT and
DBH on the tree level?

I have read in Pinheiro and Bates about the outer factors, but am not sure how
to apply it here.  I would greatly appreciate it if some someone could suggest
the correct syntax for the lme random = and also syntax to make my data into a
groupedData object.

Many thanks in advance,

Manuela