Dear SPM community,
We have a dataset with 106 subjects, and 2 within-subject factors with 2 levels (A1, A2, B1, B2). We are interested in testing the interaction effect between A & B as well as simple effects.
We have run the 1st level to get a con image for each of our factor-levels against baseline (i.e. con_A1.nii, con_A2.nii, con_B1.nii, con_B2.nii).
Then we have set up a 2nd level flexible factorial using the "Specify Subjects" option, where for each subject we use the 4 con images (so in total we have 106 subjects * 4 images = 424 images). We have defined 3 factors: subject (independent, equal variance), factorA (dependent, equal variance) and factorB (dependent, equal variance). We have also defined the interaction effect between factorA & factorB.
When checking the results for one of the simple contrasts on the GUI, the degrees of freedom are 417. We are not sure how SPM calculates the degrees of freedom so would like to check if this number is correct? It looks like really high to us, and we are worried that our model is not taking into account that we have repeated measures (i.e. that each subject has 4 images)?
Related to this, we have seen that other SPM users also add a main effect of subject, so we would like to check if given our model we should also add this? We have tried to add it but when running the model estimation we get an error that says "Please check your data: There are no significant voxels" (during spm_est_non_sphericity, line 208). So if we need to add the main effect of subject, it would be really helpful to know why we get this error message and how we could solve it.
Thank you very much for your help!
Best,
Roser
---------
Roser CaƱigueral, PhD
Postdoctoral Research Associate
Developmental Change and Plasticity Lab
University College London
|