Dear FSL staff,
I'm running a full factorial 3X2 ANOVA and I keep on having the same message in the output:
"Errors occurred during the analysis"
Although the model is unbalanced, it contains several individuals in each of the 3x2 = 6 levels. I think the dessign matrix is OK. I started from a full matrix with redundant columns
mean P1 P2 G1 G2 G3 P1_G1 P1_G2 P1_G3 P2_G1 P2_G2 P2_G3 1 1 0 1 0 0 1 0 0 0 0 0 1 1 0 0 1 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 1 0 0 0 1 0 1 1 0 0 0 0 0 1 0 0 1 0 1 0 1 0 0 0 0 0 1 0 1 0 1 0 0 1 0 0 0 0 0 1
Where P and G are factors with 2 and 3 levels (for simplicity I only show one row per combination of levels, but there are as many rows as individuals).
Then, I have eliminated columns to obtain a full rang matrix (with 6 linearly independent columns)
mean P1 G1 G2 P1_G1 P2_G3
1 1 1 0 1 0
1 1 0 1 0 0
1 1 0 0 0 0
1 0 1 0 0 0
1 0 0 1 0 0
1 0 0 0 0 1
This is the design matrix I've used (but moving the mean column of ones to the end). I have set a contrast for each column (EV) but the last one (coding for the mean).
C1 1 0 0 0 0 0
C2 0 1 0 0 0 0
C3 0 0 1 0 0 0
C4 0 0 0 1 0 0
C5 0 0 0 0 1 0
And I've asked for three different F tests
1) One for C1 (main effect of factor P [with two levels])
2) One for C2 and C3 (main effect of factor G [with three levels])
3) One for C4 and C5 (columns coding for interaction)
I haven't centered any column as (I understand it) centered and uncentered vectors span exactly the same linear subspaces, and they are equivalent in an F-test comparing the full model with a nested submodel. Here I'm assuming that each required F-test is related to a comparison between the full model and the submodel without the columns of the contrasts included in the F-test (as it is usually the case in standard GLM theory). Is that right?
I've have used FLAME1 and Mixed OLS, with common variance estimate for all groups (Group 1 for all), leading to the same error message.
Sorry for this loooooonnngg question/explanation, and thanks very much in advance.
Raymond Salvador