Hi Mark,
Apologies for not being clear. Yes i did separate analysis for
1. Pre treatment group mean
2. Pre treatment group mean with test scores
3. Post treatment group mean
4. Post treatment group mean with test scores
5. Pre vs Post
6. Pre vs Post with test scores.
The 3 results that i mentioned in the last email was from 2, 4, 6 analysis above.
I did pre and post separately just to see whether the right areas were activated for the task and not to interrogate pre vs post differences. The correlation analysis was done separately for pre and post as well to see whether test scores are correlated with BOLD in either sessions. Your F-test suggestion maybe true, the correlation results have a cluster zscore range of 2.7 to 3.3 and this was after i used a very liberal threshold of z=1.8 and cluster p=0.05. The problem is that the paradigm is event related and there are only 6-7 events of interest or even less depending upon subject response in each of 6 runs (78 frames each)and so i had to use a liberal threshold even to get mean activation. I got much better group mean results using flobs at the first level and combining the 3 copes of basis fns at the higher level via an Ftest. I havent done the paired ttest or correlation analysis yet for basis fn model. So perhaps i may get conclusive results after i figure out a way to set it up.
-Vish
Dear Vishwadeep,
The description of the demeaning of the test scores (no multiplying by -1) and the addition
of this as one extra EV sounds fine. I do not understand what you mean by "I also added
the test scores in individual session's group mean analysis". Is this a separate analysis?
It looks like you have three separate analyses here. One for pre only, one for post only
and one for pre and post. You can ignore the first two as far as interrogating a difference
between pre and post. It is also unclear how many EVs you have in the final analysis
as your contrast finished with ... However, the lack of results for the F-test in this case is
highly surprising. Are all the correlation results only just above threshold? It can be the
case that these disappear because the two-sided threshold for the F-test is more stringent.
I had assumed from your previous email that you had done two analyses: each one a pre
vs post difference analysis, but one with the test scores and one without.
Can you please clarify what you have done.
All the best,
Mark
|