[R] Error with corCompSymm and lme fit for repeated measures

[Francois Teste]

Dear R Friends,  
I need help with an error associated with corCompSymm in an lme fit.
I am using a mixed effects model to analyze a split-plot with
repeated measures and would like to fit with the compound symmetry
correlation structure.  This problem doesn't occur when using
corAR1 or any of the other structures.  I would greatly appreciate
help on how to solve this issue.  

Here's my data, code lines, and error messages:

> PhotoRed
   TreeID Block  M    D   Time           y
1       1   Dgl s1 0.5m   june  2.91130232
2       1   Dgl s1 1.0m   june  4.92232779
3       1   Dgl s1 2.5m   june  6.17724409
4       1   Dgl s1 5.0m   june  3.34278032
5       1   Dgl s1 0.5m   july  6.30782170
6       1   Dgl s1 1.0m   july  6.98652976
7       1   Dgl s1 2.5m   july  8.80715989
8       1   Dgl s1 5.0m   july  7.19424460
9       1   Dgl s1 0.5m august  4.46399690
10      1   Dgl s1 1.0m august 10.63857941
11      1   Dgl s1 2.5m august  9.84689405
12      1   Dgl s1 5.0m august 13.15311387
13      2   Dgl s4 0.5m   june  5.37380975
14      2   Dgl s4 1.0m   june  4.05953992
15      2   Dgl s4 2.5m   june  6.53184737
16      2   Dgl s4 5.0m   june  4.85340448
17      2   Dgl s4 0.5m   july  3.77109456
18      2   Dgl s4 1.0m   july  6.64288350
19      2   Dgl s4 2.5m   july  9.74241641
20      2   Dgl s4 5.0m   july  5.10884682
21      2   Dgl s4 0.5m august  8.01357594
22      2   Dgl s4 1.0m august  5.31430680
23      2   Dgl s4 2.5m august 10.62809063
24      2   Dgl s4 5.0m august 10.30284109
25      3   Dog s1 0.5m   june          NA
26      3   Dog s1 1.0m   june  3.29442388
27      3   Dog s1 2.5m   june  1.02478107
28      3   Dog s1 5.0m   june          NA
29      3   Dog s1 0.5m   july          NA
30      3   Dog s1 1.0m   july  2.19628259
31      3   Dog s1 2.5m   july  1.11794299
32      3   Dog s1 5.0m   july          NA
33      3   Dog s1 0.5m august          NA
34      3   Dog s1 1.0m august  2.44031399
35      3   Dog s1 2.5m august  0.37264766
36      3   Dog s1 5.0m august          NA
37      4   Dog s4 0.5m   june  4.33381487
38      4   Dog s4 1.0m   june  1.34196737
39      4   Dog s4 2.5m   june  0.12930756
40      4   Dog s4 5.0m   june  6.51695980
41      4   Dog s4 0.5m   july  1.33348150
42      4   Dog s4 1.0m   july  2.81174115
43      4   Dog s4 2.5m   july  7.69379970
44      4   Dog s4 5.0m   july  6.35992462
45      4   Dog s4 0.5m august  3.66707412
46      4   Dog s4 1.0m august  0.44732246
47      4   Dog s4 2.5m august  7.75845348
48      4   Dog s4 5.0m august  5.96733668
49      6     F s1 0.5m   june  6.98640483
50      6     F s1 1.0m   june  8.14953755
51      6     F s1 2.5m   june  4.50763794
52      6     F s1 5.0m   june  6.62297344
53      6     F s1 0.5m   july 13.59516616
54      6     F s1 1.0m   july  9.31375720
55      6     F s1 2.5m   july  8.01357856
56      6     F s1 5.0m   july  3.72542256
57      6     F s1 0.5m august  2.45468278
58      6     F s1 1.0m august  2.13440269
59      6     F s1 2.5m august  0.08347478
60      6     F s1 5.0m august          NA
61      7     F s4 0.5m   june          NA
62      7     F s4 1.0m   june          NA
63      7     F s4 2.5m   june  6.03948199
64      7     F s4 5.0m   june  5.73051868
65      7     F s4 0.5m   july          NA
66      7     F s4 1.0m   july          NA
67      7     F s4 2.5m   july 10.27613353
68      7     F s4 5.0m   july  8.00803252
69      7     F s4 0.5m august          NA
70      7     F s4 1.0m august          NA
71      7     F s4 2.5m august 10.27613353
72      7     F s4 5.0m august  7.34681883

> lme1 = lme(y ~ M*D*Time, random=~1|Block/M/D/Time, data=PhotoRed, na.action=na.omit)
> lme2 = update(lme1, correlation=corAR1())
> lme3 = update(lme1, correlation=corCompSymm())
Error in corMatrix.corCompSymm(object) : NA/NaN/Inf in foreign function call (arg 1)
> cs = corCompSymm(value=0.5,form=~1|TreeID)
> lme3 = update(lme1, correlation=cs)
Error in lme.formula(fixed = y ~ M * D * Time, data = PhotoRed, random = ~1 |  : 
	Incompatible formulas for groups in "random" and "correlation"
>
--
Francois Teste (Ph.D. candidate)
The University of British Columbia
Department of Forest Sciences
3041-2424 Main Mall
Vancouver, B.C., Canada, V6T 1Z4