Dear Felix,
On Thu, 10 Jun 2021 18:05:24 +0100, Felix Schmidt <[log in to unmask]> wrote:
>Dear experts,
>
>I am interested in running voxel- and surface-based morphometry analyses on two groups (placebo vs. treatment) and two timepoints (baseline vs. post-treatment) with a mixed measures ANOVA design (flexible factorial design in SPM12). I would like to investigate several pre-defined regions of interest and apply threshold free cluster enhancement method (TFCE).
Instead of applying several and separate analyses for predefined ROIs I would suggest combining these regions to only one large ROI and only apply TFCE for that region. Otherwise, you are doing independent analyses which should be finally corrected by the number of tests (e.g. Bonferroni).
>
>TFCE toolbox for SPM12 can be used on an " existing 2nd-level Model from a previous (parametric) SPM12 analysis of volume or surface data". Thus, I specified and estimated the model accordingly to the CAT12 manual instructions on pp. 40 within the SPM12 GUI.
>
> In short:
>- Factors: subject, group, time
>- main effect: subject [1]
>- interaction: group x time [2 3]
>- covariates: age, tiv, bmi
I have to urgently add this to the CAT12 manual. Covariates that don't change over time should be not used in a longitudinal design! There are only some rare scenarios where it makes sense to add covariates to a flexible factorial design for longitudinal data. Please check the SPM mail list for use of covariates in flexible factorial designs. This is a re-occurring issue.
>- factor matrix:
>
>Scan x Subject x Group x Session
>
> 1 1 1 1
>
> 1 1 1 2
>
> 1 2 1 1
>
> 1 2 1 2
>
> ….
>
>- contrasts for interactions: [1 -1 -1 1] and [-1 1 1 -1]
>- statistical design is attached to this post
>
>Unfortunately, the SPM12 model estimation fails for only 3 out of 11 masks (all created under the same constraints with wfu pickatlas tool), printing out the error message:
>
>10-Jun-2021 18:58:49 - Running 'Estimate TFCE'
>10-Jun-2021 18:58:56 - Failed 'Estimate TFCE'
>Error using spm_conman (line 476)
>Contrast Manager was quit!
>In file "C:\Program Files\spm12\spm12\spm_conman.m" (v7738), function "spm_conman" at line 476.
>In file "C:\Program Files\spm12\spm12\toolbox\TFCE\tfce_estimate_stat.m" (v219), function "tfce_estimate_stat" at line 124.
>
>The following modules did not run:
>Failed: Estimate TFCE
>
>The exact same model estimation succeeded for the rest of the masks, but it failed only for anterior and posterior cingulate and fusiform gyrus regions.
>
>What could be the reason for this error message and how could it possibly be solved?
Please repeat the analysis with my suggestions. Focus on one analysis with one ROI combining your several regions and remove covariates from your design. And cross fingers that it runs smoothly...
Best,
Christian
>
>Thank you very much in advance! Your help would be very appreciated!
>
>With kind regards
>
>Felix
>
|