I have been reading messages on the listserve about entering covariates of interest in ANOVAs, that make me think that this is not ok. I need to examine covariates of interest in relation to two different data sets, both of which use a faces (vs shapes) task. Could you please advise on whether it is ok to use covariates in the following ways:
1. Relating BOLD and a peripheral psychophysiological measure: I have a 2(Group: A, B) X 3(Condition: Angry, Fearful, Happy) full factorial ANOVA. To examine how a psychophysiological variable recorded simultaneously covaries with BOLD and how this might differ by group, condition or groupXcondition, I entered two covariates:
a) Group A's psychophysiological measure (ordered by condition then subject: Angry subject 1, Angry subject 2...Fearful subject 1, Fearful subject 2...Happy subject 1, Happy subject 2...), with the rows for Group B padded with zeros
b) Group B's psychophysiological measure (ordered by condition then subject: Angry subject 1, Angry subject 2...Fearful subject 1, Fearful subject 2...Happy subject 1, Happy subject 2...), with the rows for Group A padded with zeros
I stipulated that each covariate interacted with Condition, so in the final model, there were 6 covariate columns (Group A Angry, Group A Fearful, Group A Happy, Group B Angry, Groups B Fearful, Group B Happy).
After running the ANOVA, I examined the main effect of condition for the covariate (i.e., same F test as is produced by SPM for examining condition effects on BOLD, but shifted over to be above the columns corresponding to the covariates). I also examined the groupXcondition covariate effect (again, same F test as one would normally use to examine groupXcondition effects, but shifted over so that it was in the columns corresponding to the covariates).
2. Full factorial ANOVA with one factor (Condition: Angry, Fearful, Happy). I am interested in knowing how BOLD within or across conditions relates to two individual difference variables, so I would like to add them as covariates of interest, however I am not sure how to do this. I get row mismatch errors if I simply enter each covariate once, so I am thinking I need to copy and paste the same covariate 3 x in the column? But that seems odd. Please advise.
Thanks in advance - your help would be very much appreciated.
- AM
|