[R] ANCOVA with defined error terms

[Richard M. Heiberger]

The three-way interactions you mention are included in the model formula
I suggested.  If they didn't appear in the expansion, it suggests
that you have some aliasing due to empty cells.

I can't do any more without your dataset.
You can post your dataset with random response values.
The exact data.frame for the predictors is needed.
Anything that is a factor must be a factor in what you send.
You can use dput to ensure accurate copying.


 > tmp <- data.frame(a=factor(rep(letters[1:3], 2)), y=rnorm(6))
 > tmp
   a          y
1 a -0.4313252
2 b -0.9065241
3 c -0.7285257
4 a  0.0368019
5 b  0.7982373
6 c -2.4712612
 > dput(tmp)
structure(list(a = structure(c(1L, 2L, 3L, 1L, 2L, 3L), .Label = c("a",
"b", "c"), class = "factor"), y = c(-0.431325155041393, -0.906524086347679,
-0.728525691910586, 0.0368018971284965, 0.798237317168781, -2.47126116272324
)), .Names = c("a", "y"), row.names = c(NA, -6L), class = "data.frame")
 >
 >
 > structure(list(a = structure(c(1L, 2L, 3L, 1L, 2L, 3L), .Label = c("a",
+ "b", "c"), class = "factor"), y = c(-0.431325155041393, 
-0.906524086347679,
+ -0.728525691910586, 0.0368018971284965, 0.798237317168781, 
-2.47126116272324
+ )), .Names = c("a", "y"), row.names = c(NA, -6L), class = "data.frame")
   a          y
1 a -0.4313252
2 b -0.9065241
3 c -0.7285257
4 a  0.0368019
5 b  0.7982373
6 c -2.4712612
 >