Hi, 
I have 24 conditions from first level analysis , the first 8 are response to stimuli#1 and second 8 stimuli#2 and last 8 to stimuli#3. In order to do PPI analysis and compare stimuli effect I put the first 8 on the left and second 8 on the right side to form P but the result of this is very different from when I merged this 8 conditions from the beginning and run first level for 3 conditions(stimuli) . I wonder if there is any explanation or recommendation to this kind of analysis. The reason of having 24 conditions is that we add more details to be more flexible in creating contrasts. for example stimuli#1, hard, positive, level1 and so on.
BR

Hi,

You have created a model with a lot of collinearity (high correlations) between regressors. PPI regressors for conditions with low trial numbers are very similar to the task regressors. This leads to extremely unstable estimates of both the task and PPI terms.

I would recommend not splitting each stimuli into 8 types unless you had many stimuli of each type (>10).

I would also not recommend a model with 8 conditions for one stimulus type at the first level.


Best,
Donald


Originally posted by Amirhossein Manzouri:

Hi Donald,
Thanks for the prompt and informative reply .
BR