Hi Anderson,
Thanks for taking the time to look at this problem! The clarification on the website is very helpful. However, I restarted the analysis with the new PALM version and now encounter the following error:
error: out of memory or dimension too large for Octave's index type
error: called from:
error: /usr/local/LKeb/FSL/PALM65/palm_core.m at line 616, column 32
error: /usr/local/LKeb/FSL/PALM65/palm.m at line 80, column 1
This is probably not a memory problem, as it even shows up with 1 input and 1 EV (there are no errors when I use 10 inputs without EVs). We are not sure how to solve this. No matter how many input/EVs I use and how much memory I assign to the job, it gets stuck at the same point:
….
Reading design matrix and contrasts.
Elapsed time parsing inputs: ~ 15.7903 seconds.
Number of possible permutations is 1.84606e+55.
Generating 100 shufflings (permutations only).
Building null distribution.
Doing maths for -evperdat before model fitting: [Design 1/1, Contrast 1/2] (may take some minutes)
Hopefully, you know how to overcome this.
Furthermore, with regard to my design: I added an excel file with the mixed effects design and contrasts as currently set up. We modelled treatment and time as fixed factors plus random intercepts for the (12) different subjects. There are 2 study days, on one day the subjects receive a drug and on the other a placebo. On both days, 1 baseline scan was made (pre dosing) and after receiving drug or placebo, 4 additional (post dosing) scans were acquired. We are mainly interested in the drug treatment effect —> for each of 10 modalities (with equal resolution), is connectivity decreased and/or increased as a consequence of taking drug vs. placebo? The plan is to add the baseline scan (Z-map) as voxelwise EV, to correct for possible differences at baseline level (instead of subtracting baseline from each post scan).
Additional analyses of interest might be:
- Treatment effect for each time point separately
- Treatment x time interaction effect
MAN(C)OVA or NPC may be possible as well, although we are interested in the effects for each modality separately, which will probably lead to subsequent post-hoc tests. What is your suggestion? I would be happy to receive your advice about the appropriate design/analysis.
Best,
Bernadet
|