Hi,
I have a dataset of 8 subjects, who were each scanned on between 13 and 50 occasions. I am interested in the impact of a couple of DAYILY measures that vary between the different scan days.
At the first level, I will enter things like motion correction time series for each scan, and, if it were one subject, I would enter my DAILY variables at the second level. But how should this be done for multiple subjects?

Can I enter 8 "DAILY" variables at the second level, where each subject gets her numbers, and the rest are coded as zeros?
Or should I model the effect of DAILY in each subject in a separate GLM, and then do a third level test in SPM using the resulting con*.img maps? And if so, how do I weight for the number of scans each subject underwent?

Best
Clas

Hi Clas,

Yes, I would do exactly the latter, modeling the effect of DAILY in each subject in a separate GLM and then entering the resulting model parameters into a new second-level analyses (now across subjects). We are working to incorporate a way to facilitate exactly this form of 3-levels analysis into CONN hopefully for the next release but until that is ready one relatively simple way to do this would be (I am assuming here from your description that you are entering each scan separately as if it was a different subject, please let me know otherwise)

1) create 8 second-level covariates named S1, S2, ... S8, containing 1's for the scans associated with each subject, and 0's otherwise

2) create 8 second-level covariates named D1, D2, ... D8, for your DAILY measures containing the actual values for each subject (and 0's for the other subjects)

3) run one second-level analysis per subject, in the first one selecting S1 and D1 and entering a [0 1] contrast, in the second one selecting S2 and D2, etc.

4) run a third-level analysis selecting the 8 con0001.nii images from the above 8 results and performing a one-sample t-test, perhaps using something like:

  X = ones(8,1); %design matrix (subjects-by-effects matrix; 8x1)
  Y = conn_dir('/data/myconnresultsfolders/con*.nii','-cell'); % data (subjects by conditions cell array; 8x1)
  C = 1; % between-subjects contrast
  M = 1; % between-conditions contrast
  pathname = '/data/thirdlevelresults'; % output folder for third-level SPM.mat and associated files
  conn_module('glm',X,Y,C,M,pathname);

Hope this helps
Alfonso
Originally posted by Clas Linnman:

Thanks!
C