Hello!

I have a partially longitudinal resting fMRI dataset (participants visited once or twice). In each session, participants completed a resting-state scan in which they saw either a fixation cross or a movie. This was not necessarily a between-subjects variable, because some participants saw fixation in session 1 and movie in session 2, or vice versa. I'd like to perform a contrast between fixation and movie.

To do so, I have created 4 covariates to capture the fact that one participant may have seen a movie on their first visit, but a fixation on their second:

• Fixation_01 (coded as 1 if they viewed a fixation cross in session 1, 0 if they did not, NaN if there is missing data; 133 non-NA subjects)


• Movie_01 (coded as 1 if they viewed a movie in session 1, 0 if they did not, NaN if there is missing data; 133 non-NA subjects)


• Fixation_02 (same as above but for session 2, 43 non-NA subjects)


• Movie_02 (same as above but for session 2, 43 non-NA subjects)

CONN also automatically sorted sessions into the "conditions" tab so I have the conditions of 01_rest and 02_rest.

Now, I would like to contrast fixation vs. movie averaging across both sessions to see if there is a difference, to do this I tried running:

between-subjects contrast: [1 -1 1 -1] (with the variables in the order I listed above)

between-conditions contrast: [1/2 1/2]

However, I get this warning and a design matrix with only 37 samples (I am unclear on where this number is coming from).

Is there something wrong in the way I have set up the covariates and/or the contrast? 

Thank you so much for your help!!