multisession glm means running all of your sessions in one model. I don't know how to do it in the gui because i have all my analysis scripted, but i can't imagine its too hard to figure out.

while one could argue that their maybe scan specific baseline activity across sessions, the 'constant' regressor theoretically should absorb the noise, which can facilitate comparisons across sessions.

cheers,
michael

On Thu, Jan 13, 2011 at 11:31 PM, Woogul Lee <[log in to unmask]> wrote:

Dear,

Thank you for your answer.

I think I need to explain my study design more specifically.

We have three runs.

In the first run, there were trials of the baseline condition only.

In the second run, there were trials of the "A" condition only.

In the third run, there were trials of the "B" condition only.

Now, we tried to compare the activation differences between A and baseline, between B and baseline, or between A and B. 

Are there ways that we can do that?

You mentioned about the multisession glm.

In this situation, we are wondering whether the standardized beta coefficient of the A condition, that of the B condition, and that of baseline are comparable.

As beta coefficient of each condition was calculated based on the mean values of each target run (first run for baseline, second run for A condition, third run for B condition), we think the standardized beta values are not comparable. 

Could you let me know whether we still need to consider this problem in the multisession glm that you suggested, what the multisession glm is exactly, and how we conduct it with SPM?

Thanks.

Woogul Lee

On Fri, Jan 14, 2011 at 2:50 PM, Michael T Rubens <[log in to unmask]> wrote:

not sure I fully  understand you're problem, i fail to see an issue. what is the nature of your conditions?

Just run a multisession glm, which will pop out 3 betas, and then just make your contrasts.

Cheers,
M

On Thu, Jan 13, 2011 at 9:16 PM, Woogul Lee <[log in to unmask]> wrote:

Dear all

 

I am stuck in the design problem so I sincerely hope you to help me. 

 

Basically, our design is block design consisting of three DIFFERENT runs. The problem is that each run has only one condition (sigh…) so I cannot compare A condition with B or C condition directly. Unfortunately, I realized the fact very lately after conducting the experiment.

 

Is there any way I can compare those different conditions statistically??

 

If you have any idea, please, help me.

 

 

Best regards

 

Woogul Lee

--
Research Associate
Gazzaley Lab
Department of Neurology
University of California, San Francisco

--
Research Associate
Gazzaley Lab
Department of Neurology
University of California, San Francisco