I see this is a popular question -- and have tried to apply the help given
to other folks, but I keep hitting a wall...
I've used the 2x2 ANOVA design in the FEAT manual with success. (my model =
Gender x Stimulus type (2 types, A and B); 4 subjects for sake of argument).
If I assume independent measures for both factors and use the example model
(below, a 4 sj example of the model I used) I encounter no errors:
EV1 EV2 EV3 EV4
input COPE Group Stim Gender Interaction Mn
Female_sj1_stimA 1 -1 -1 1 1
Female_sj2_stimA 1 -1 -1 1 1
Male_sj3_stimA 1 -1 1 -1 1
Male_sj4_stimA 1 -1 1 -1 1
Female_sj1_stimB 1 1 -1 -1 1
Female_sj2_stimB 1 1 -1 -1 1
Male_sj3_stimB 1 1 1 1 1
Male_sj4_stimB 1 1 1 1 1
The issue is, as you may have noticed -- my group is not 8 independent
subjects, rather 4 subjects with a repeated measure (stimulus) and an
independent factor (gender). I'd like to model out within subject
variability: i.e. like the paired t-test.
I see this was recommended to several people asking a similar question.
This model looks like:
EV1 EV2 EV3 EV4 EV5 EV6 EV7
inputCOPE Group Stim Gender Interaction sj1 sj2 sj3 sj4
Fem_sj1_stimA 1 -1 -1 1 1 0 0 0
Fem_sj2_stimA 1 -1 -1 1 0 1 0 0
Male_sj3_stimA 1 -1 1 -1 0 0 1 0
Male_sj4_stimA 1 -1 1 -1 0 0 0 1
Fem_sj1_stimB 1 1 -1 -1 1 0 0 0
Fem_sj2_stimB 1 1 -1 -1 0 1 0 0
Male_sj3_stimB 1 1 1 1 0 0 1 0
Male_sj4_stimB 1 1 1 1 0 0 0 1
I assume that this accounts for the dependence of inputs 1 and 5, 2 and 6,
and so on.
Problem I run into is everytime I try and use this model it reports a rank
deficient error.
Any ideas on what might I be missing?
Thank you,
~Jonathan
|