Dear Anderson,

Thank you so much for your help and the clever solution proposed. I tried to implement your design with the 3 EVs. As you indicated EV3 has only ones sinces EV1 and EV2 model within-subject differences anyway:

EV1: A column that is the following vector transposed:
[1 -1 0  1 -1 0  1 -1 0  1 -1 0 ... 1 -1 0]

EV2: A column that is the following vector transposed:
[1 0 -1 1 0 -1  1 0 -1  1 0 -1 ...  1 0 -1]

EV3: all ones
[1 1 1 1 1 1  1 1 1   1 1 1 ... 1 1 1]

groups, contrasts and F-tests also implemented as advised. However, this design produced an error: "This (f)contrast appears to be rank defficient,"

Perhaps this is due to the fact that EV3 is not modelled? So I tried:

EV1: A column that is the following vector transposed:
[1 -1 0  1 -1 0  1 -1 0  1 -1 0 ... 1 -1 0]

EV2: A column that is the following vector transposed:
[1 0 -1 1 0 -1  1 0 -1  1 0 -1 ...  1 0 -1]

EV3: A column that is the following vector transposed:
[0 1 -1 0 1 -1 0 1 -1 0 1 -1 ... 0 1 -1]

EV4: all ones
[1 1 1 1 1 1  1 1 1   1 1 1 ... 1 1 1]

groups, contrasts and F-tests again implemented as advised. This design did not produce a rank defficient error. It this design correct? The F value seem to make sense. The t-values of the contrast not, very small values and they do not show the effects I earlier found using the one-sample t-test approach on the separate cope values. Can you confirm this is correct?

Unfortunately, I also discovered that the -c 2.3 option I used for all the other contrasts in my analyses seem not to work for the ftest. In other words, there is no clustere_corrp_fstat1.nii.gz file! How can I ask randomise to produce the cluster-based thresholding output for ftests? It seem to only produce voxelwise and TFCE corrected output for ftests.

Thanks in advance for your help.

Best regards, Henk