[R-sig-ME] nlmer model specification

[James Widman]

Happy Holidays to All,

  I'm trying to use nlmer to model some data from an experiment I ran 
and am having some trouble specifying the model correctly. Would 
appreciate any insight you can provide.

I am examining the growth of fish fed two diets over 9 weeks - with 
weekly measurements. Thirty fish were placed in each of 8 tanks, 4 tanks 
fed foodtype NRD and 4 tanks fed foodtype OTO.
Unbalanced data due to mortality.
I'm trying to fit a nlmer to the data and want to make sure I have 
specified the model correctly. I can provide additional information if 
needed.
This is what I came up with and have tried a number of alternatives but 
don't seem to  be progressing.  I realize the model will probably be 
simplified.

foodtype should be a fixed effect.

 > str(scupgrowth)
'data.frame':   2925 obs. of  11 variables:
  $ StartDate  : Factor w/ 1 level "2008/07/29": 1 1 1 1 1 1 1 1 1 1 ...
  $ SampleDate : Factor w/ 10 levels "2008/07/29","2008/08/05",..: 1 1 1 
1 1 1 1 1 1 1 ...
  $ FishID     : int  1 2 3 4 5 6 7 8 9 10 ...
  $ Length..mm.: int  38 39 34 33 36 40 39 34 32 35 ...
  $ Weight..g. : num  0.64 0.81 0.47 0.48 0.62 0.78 0.7 0.47 0.47 0.62 ...
  $ Tank       : int  1 1 1 1 1 1 1 1 1 1 ...
  $ Days       : int  0 0 0 0 0 0 0 0 0 0 ...
  $ FishIDf    : Factor w/ 2925 levels "1","2","3","4",..: 1 2 3 4 5 6 7 
8 9 10 ...
  $ foodtype   : Factor w/ 2 levels "NRD","OTO": 1 1 1 1 1 1 1 1 1 1 ...
  $ tankf      : Factor w/ 8 levels "1","2","3","4",..: 1 1 1 1 1 1 1 1 
1 1 ...
  $ week       : num  0 0 0 0 0 0 0 0 0 0 ...
 > (nlm1 <- nlmer(Weight..g. ~ SSgompertz(week, Asym, a, b) ~ foodtype + 
tankf + tankf|FishIDf,
+  scupgrowth, start = c(Asym = 72.70484, a = 5.34048, b = 0.83792)))
Nonlinear mixed model fit by the Laplace approximation
Formula: Weight..g. ~ SSgompertz(week, Asym, a, b) ~ foodtype + tankf 
+      tankf | FishIDf
    Data: scupgrowth
   AIC  BIC logLik deviance
  6010 6249  -2965     5930
Random effects:
  Groups   Name        Variance   Std.Dev.  Corr
  FishIDf  foodtypeOTO 4.7489e-06 0.0021792
           tankf2      2.8193e-04 0.0167909 -0.829
           tankf3      1.4213e-04 0.0119219  0.001  0.110
           tankf4      2.0968e-04 0.0144804 -0.913  0.759  0.056
           tankf5      1.8160e-04 0.0134760 -0.023 -0.007 -0.035  0.019
           tankf6      3.7620e-04 0.0193958 -0.800  0.704  0.014  0.731 
-0.260
           tankf7      2.0907e-04 0.0144592  0.023 -0.195 -0.062 -0.022 
-0.038
           tankf8      3.4983e-04 0.0187038 -0.958  0.701 -0.033  0.870  
0.222
  Residual             1.1874e+00 1.0896851

  -0.021
   0.702 -0.107

Number of obs: 2925, groups: FishIDf, 2925

Fixed effects:
       Estimate Std. Error t value
Asym 90.150991   5.205204   17.32
a     5.160145   0.042640  121.02
b     0.868886   0.003198  271.73

Correlation of Fixed Effects:
   Asym  a
a 0.748
b 0.970 0.576

When I tried the following I get an error.
   (nlm1 <- nlmer(Weight..g. ~ SSgompertz(week, Asym, a, b) ~ (1 | 
foodtype) + tankf + tankf|FishIDf,
+  scupgrowth, start = c(Asym = 72.70484, a = 5.34048, b = 0.83792)))
Error in model.matrix.default(eval(substitute(~expr, list(expr = 
x[[2]]))),  :
   model frame and formula mismatch in model.matrix()

Thanks for any feedback you can provide.
Jim Widman