 |
 |
Dear SPMers, I have 2 fmri scan runs, one done after the other, without the participant leaving the scanner. The first scan run has 5 conditions randomised in a particular order. The second scan run has 5 similar conditions (but with a constant auditory input added throughout) randomised in the same order as the first scan run. The above is then repeated at a later time point. The TR is 3. Before every condition is a brief instruction lasting 3 seconds (1 image). Each condition lasted 21 seconds (7 images). 1) Thus my design matrix has 5 +1 +5 + 1 + 5 + 1 + 5 + 1 = 24 columns for each participant. Is this correct or can I not put all 4 scan runs (at 2 time points) in the same matrix? 2) During model specification, I put the 960 images (240 per scan run) together in a folder and keyed in 3 for 'interscan interval'. There are 2 options for 'units for design': 'scans' or 'seconds'. I picked 'scans'. For each condition, I set the 'onset' as the position of the image in the folder (e.g. '1' for the first instruction (image number 1), '2' for the first condition (image number 2-8) followed by '9' for the second instruction (image number 9) and so on). For each instruction, I set the 'duration' as '1' (i.e. 1 image from onset) and for each condition, I set the 'duration' as '7' (i.e. 7 images from onset). Are my interpretation and input of 'scans', 'onsets' and 'durations' correct? 3) During model specification, there is a 'global intensity normalisation' section with 2 options: 'scaling' or 'none'. I notice that whether I pick 'scaling' or 'none', the design matrix that spm churns out will always have 'grand mean scaling: session specific'. What is 'global intensity normalisation' and 'grand mean scaling'? Since spm uses 'grand mean scaling: session specific' no matter what, does it mean I should select 'none' for 'global intensity normalisation'? Do I select 'scaling' or 'none' for 'global intensity normalisation'? 4) After model specification, spm suggests checking it using the a) design matrix b) explore design matrix and c) design orthogonality functions. What do I check for each specifically? 5) Now that I have put 4 scan runs (at 2 time points) in the same matrix, my understanding is that I can compare between the 5 conditions within each scan run without additional concern. Can I also compare between the 2 scan runs, done one after another on same day (e.g. condition 1 in first scan run and condition 1 (same but with a constant auditory input) in the second scan run)? Can I also compare between time points (e.g. condition 1 in first scan run during first time point and condition 1 in first scan run during second time point)? I added the rp file created during realignment of the 960 images (i.e. estimated time series of translations and rotations) under 'multiple regressors'. Do I need to take any additional steps? 6) Most of my contrasts would have a vector sum of 0. I know that [-1 -1 2 0 0 0 ] tests if the activity in 2 is larger than the mean activity in -1 and -1. However one particular contrast I have is e.g. [-1 -1 1 0 0 0] to test if the activity in 1 is larger than the sum of activity in -1 and -1. The activity in 1 is about double that of -1. Is [-1 -1 1 0 0 0] correct? It will be great if anyone can share papers/resources with similar contrasts. Any advice is much appreciated! Many thanks, PLC