Hi Rick,
Thanks for the feedback.
The experiment is auditory, so I'll use Silent/Listening to describe the trials.
The design is based on two manipulations such that one of the two first conditions must be used during all stim blocks. So, if the first manipulation is condition A or B (and I'm interested in the difference between A & B) and the second manipulation is condition 1, 2, 3, 4 (and I'm interested in 1 vs. 2), then the blocks are:
A1, A2, A3, A4, B1, B2, B3, B4
So, A and B must necessarily be mirror images of each other, aside from Silent blocks.
I realize the amount of Silent blocks (0's on all regressors) in between Listening blocks is low . Is that the problem?
I do have a long Silent block at the end. If it were longer, would that help?
If not, is there anything I can do? I believe in BrainVoyager there is a way of multiplying the different conditions so that it's balanced (i.e. multiply Silence x 8) and I do actually get the main effect of just (A - baseline) that I'd expect using BV.
The last regressor is my sanity check, making sure there is a consistent effect of just Listening vs. Silence - it's the sum A1+A2+...+B4. So it's "1" 8/9ths of the time aside from a long Silent block at the end. If I removed this regressor would that fix the analysis?
Thank you,
Marc
tester wrote:
> Hi Marc,
>
> Those 5 regressors are highly correlated. You can see it
> visually when using just "1dplot -sepscl my.xmat.1D" (if
> we suppose your X-matrix is called my.xmat.1D). You can
> get the correlation values with 1d_tool.py, too:
>
> 1d_tool.py -infile my.xmat.1D -show_cormat_warnings
>
>
> That requires the latest AFNI binaries, from last night.
>
> ----
>
> In particuar, regressors 17 and 18 (your first 2 of interest)
> are practially just negatives of each other.
>
> On top of that, regressor 21 (the last one) is on practically
> the entire time. These will certainly start to match the
> baseline when they are put together.
>
> ----
>
> It's a little difficult to ponder without knowing your design
> in detail. But the curves look problematic, when taken
> together.
>
> Also, for a block design, I do not think it is good to have
> multiple stimulations at the same time. If a brain location
> has already plateaued/saturated due to one stimulus, I do not
> expect much difference to be seen from another stimulus that
> is supposed to have a response plateau at the same time.
>
> - rick
>
> P.S. You do not need to contact Bob separately. It is just
> good to tell the person that you are communicating with, so
> they know the data has been uploaded.
>