Dear Jon,
> Help appreciated. I have a simple task with 2 conditions (A,B), for each individual at the first level I have contrasted A vs B, at the second level I have ran a two-sample t-test on the A vs B contrast with assumption that this would give me the group by task interaction.
I am assuming that the subjects fall into two groups? If so, yes a two-sample t-test where each sample consists of the A-B copes for one group will give you the group-by-task interaction.
> However, when I extract cope values for A and B seperately (data extracted from the mask that was significant for the A vs B contrast) and put these through SPSS etc the interaction is not significant - although, of course, if I use these values to calculate A-B then the two-sample t-test is significant. I have seen a few posts recently suggestng that a two-sample t-test on the the first level difference images (A vs B) gives the group x task interaction
When you set up the SPSS analysis do you specify that the paired A and B belong to the same subject and ask for a "blocked" analysis (I am not sure if this is SPSS jargon, but you get my drift)? If not it would not be terribly surprising if it doesn't reach significance.
Jesper
|