Dear Yann,
On 8 Feb 2019, at 8:27, Yann Quidé wrote:
> Dear Christian,
>
> Thanks for your reply. I selected all atlases for preprocessing; so I
> should be able to select any atlas. I was able to use this atlas for
> another set of analysis, but not this one.
>
> Any idea why? Or maybe I should change something in the default
> parameters?
If you select the atlases in the GUI this should be sufficient. However,
if this not worked for you the change in the defaults is the safest way.
I will check that the GUI selection will be really used.
Best,
Christian
>
> Thanks for your time,
>
> Yann
>
>
>> On 8 Feb 2019, at 6:16 pm, Christian Gaser
>> <[log in to unmask]> wrote:
>>
>> Dear Yann,
>>
>> On Thu, 7 Feb 2019 16:43:28 +1100, Yann Quidé <[log in to unmask]>
>> wrote:
>>
>>> Dear all,
>>>
>>> 1) Thanks Guillaume.
>>> Just for reference, when entering manually the contrast in the
>>> contrast manager (after building and estimating the model), it
>>> crashes (as you mentioned), but not when I enter [ones(1,2)/2
>>> -ones(1,2)/2 0 ones(1,15)/15 -ones(1,10)/10].
>>> In addition, when looking at the results for this contrast, there is
>>> a significant main effect of group all over the place; which is
>>> definitely not possible in my population.
>>> Anyway I will try your option.
>>>
>>>
>>> 2) I have another question.
>>> When using the CAT12 option for ROI analyses, and comparing Time 1
>>> vs Time 2 from my longitudinal design, I do not have the choice of
>>> the atlas; CAT12 automatically selects the neuromorphometrics atlas
>>> (for volumetric measures). Is there any reason for that? How can I
>>> use another atlas for this analysis (eg Cobra)?
>> You need to select the used atlases before preprocessing. It's not
>> possible to do that retrospectively.
>>
>> Best,
>>
>> Christian
>>>
>>> Thanks for your time!
>>>
>>> Yann
>>>
>>>
>>>
>>>
>>>> On 18 Jan 2019, at 10:33 PM, Flandin, Guillaume
>>>> <[log in to unmask]> wrote:
>>>>
>>>> Dear Yann,
>>>>
>>>> Could it be that you entered the contrast as:
>>>> [0.5 0.5 -0.5 -0.5 0 0.0667 ... ]
>>>> instead of:
>>>> [[1 1 -1 -1]/2 0 ones(1,15)/15 -ones(1,10)/10]
>>>> so that the contrast vector does not sum to zero (at machine
>>>> precision)?
>>>>
>>>> Again, for the main effect of group, I would use a two sample
>>>> t-test
>>>> after having averaged the two images per subject.
>>>>
>>>> Best regards,
>>>> Guillaume.
>>>>
>>>>
>>>> On 17/01/2019 22:48, Yann Quidé wrote:
>>>>> Thanks Guillaume and Christian,
>>>>>
>>>>> Christian, I tried without TIV as covariate, but got the same
>>>>> error.
>>>>>
>>>>> I have attached the related SPM.mat, the Error.mat saved following
>>>>> the
>>>>> crash, and the details from the Matlab command window.
>>>>>
>>>>> Please let me know if you need anything else.
>>>>>
>>>>> Thanks again for your time and advice on this.
>>>>>
>>>>> Cheers,
>>>>>
>>>>> Yann
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------
>>>>> Running job #1
>>>>> ------------------------------------------------------------------------
>>>>> Running 'Factorial design specification'
>>>>> Mapping files :
>>>>> ...done
>>>>> Design configuration :
>>>>> ...done
>>>>> Saving SPM configuration : ...SPM.mat
>>>>> saved
>>>>> Design reporting :
>>>>> ...done
>>>>> Completed : 09:43:00 -
>>>>> 18/01/2019
>>>>> Done 'Factorial design specification'
>>>>> Running 'Model estimation'
>>>>>
>>>>> SPM12: spm_spm (v7120) 09:43:00 -
>>>>> 18/01/2019
>>>>> ========================================================================
>>>>>
>>>>> SPM12: spm_est_non_sphericity (v6913) 09:43:00 -
>>>>> 18/01/2019
>>>>> ========================================================================
>>>>> Chunk 2/2 :
>>>>> ...done
>>>>> Temporal non-sphericity (70376 voxels) : ...ReML
>>>>> estimation
>>>>> ReML Iteration 1 : ...2.662013e-01
>>>>> [+4.25]
>>>>> ReML Iteration 2 : ...2.289632e-04
>>>>> [+4.50]
>>>>>
>>>>> SPM12: spm_spm (v7120) 09:43:05 -
>>>>> 18/01/2019
>>>>> ========================================================================
>>>>> Chunk 2/2 :
>>>>> ...done
>>>>> Spatial non-sphericity (over scans) :
>>>>> ...done
>>>>> Saving SPM.mat :
>>>>> ...done
>>>>> Completed : 09:43:30 -
>>>>> 18/01/2019
>>>>> Done
>>>>> Done 'Model estimation'
>>>>> Running 'Contrast Manager'
>>>>>
>>>>> SPM12: spm_contrasts.m 09:43:30 -
>>>>> 18/01/2019
>>>>> ========================================================================
>>>>> Contrasts folder :
>>>>> ..tudinal_Time_by_Group_no_age
>>>>> 0.5 0.5 -0.5 -0.5 0 0.066667
>>>>> 0.066667 0.066667 0.066667 0.066667 0.066667
>>>>> 0.066667
>>>>> 0.066667 0.066667 0.066667 0.066667 0.066667
>>>>> 0.066667
>>>>> 0.066667 0.066667 -0.1 -0.1 -0.1
>>>>> -0.1
>>>>> -0.1 -0.1 -0.1 -0.1 -0.1
>>>>> -0.1
>>>>> <- !invalid contrast
>>>>> Failed 'Contrast Manager'
>>>>> Error using spm_run_con (line 251)
>>>>> Error in contrast specification
>>>>> In file
>>>>> "/mridata/workingdata/RUSE/spm12_coptsd_cat12.5/config/spm_run_con.m"
>>>>> (v7093), function "spm_run_con" at line 251.
>>>>>
>>>>> The following modules did not run:
>>>>> Failed: Contrast Manager
>>>>>
>>>>>> On 15 Jan 2019, at 8:38 AM, Christian Gaser
>>>>>> <[log in to unmask] <mailto:[log in to unmask]>
>>>>>> <mailto:[log in to unmask]
>>>>>> <mailto:[log in to unmask]>>> wrote:
>>>>>>
>>>>>> Dear Yann,
>>>>>>
>>>>>> On Fri, 11 Jan 2019 18:17:57 +1100, Yann Quidé
>>>>>> <[log in to unmask] <mailto:[log in to unmask]>
>>>>>> <mailto:[log in to unmask] <mailto:[log in to unmask]>>> wrote:
>>>>>>
>>>>>>> Hi all,
>>>>>>>
>>>>>>> Any thought on this?
>>>>>>
>>>>>> To be honest, not. The designs are fine and the contrast for the
>>>>>> 2nd
>>>>>> model with the group differences is also correct. The only idea I
>>>>>> have
>>>>>> is to skip the TIV covariate.
>>>>>>
>>>>>> Best,
>>>>>>
>>>>>> Christian
>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>> Yann
>>>>>>>
>>>>>>>> On 9 Jan 2019, at 9:08 pm, Yann Quidé <[log in to unmask]
>>>>>>>> <mailto:[log in to unmask]>
>>>>>>>> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>> Dear all,
>>>>>>>>
>>>>>>>> I am analysing longitudinal VBM data (group1, n=10; group2,
>>>>>>>> n=15; 2
>>>>>>>> time points) using CAT12.5 (r1363).
>>>>>>>>
>>>>>>>> 1) Looking within group for the effect of time my design matrix
>>>>>>>> looks like attachment1… all boxes (but the one associated
>>>>>>>> with TIV)
>>>>>>>> below the matrix are grey, indicating that these data are not
>>>>>>>> uniquely specified. As I am comparing the “same brains”
>>>>>>>> over time, I
>>>>>>>> believe this is ok; is that right?
>>>>>>>>
>>>>>>>> 2) I also wanted to look at the main effects of time, group,
>>>>>>>> and
>>>>>>>> their interaction using a 2x2 ANCOVA design (design in
>>>>>>>> attachment2).
>>>>>>>> I followed the CAT12 manual, and used
>>>>>>>> ones(1,n)/n -ones(1,n)/n 0 ones(1,n1)/n1 -ones(1,n2)/n2
>>>>>>>> to model the main effect of group (n=2; n1=15; n2=10).
>>>>>>>>
>>>>>>>> However, the creation of the contrast image crashes when using
>>>>>>>> the
>>>>>>>> batch mode, with the following error message:
>>>>>>>>
>>>>>>>> ========================================================================
>>>>>>>> Contrasts folder :
>>>>>>>> ..tudinal_Time_by_Group
>>>>>>>> 0.5 0.5 -0.5 -0.5 0 0.066667
>>>>>>>> 0.066667 0.066667 0.066667 0.066667 0.066667
>>>>>>>> 0.066667 0.066667 0.066667 0.066667 0.066667
>>>>>>>> 0.066667 0.066667 0.066667 0.066667 -0.1
>>>>>>>> -0.1 -0.1 -0.1 -0.1 -0.1
>>>>>>>> -0.1 -0.1 -0.1 -0.1 <- !invalid
>>>>>>>> contrast
>>>>>>>> Failed 'Contrast Manager'
>>>>>>>> Error using spm_run_con (line 251)
>>>>>>>> Error in contrast specification
>>>>>>>> In file "/Users/y.quide/spm12/config/spm_run_con.m" (v7093),
>>>>>>>> function "spm_run_con" at line 251.
>>>>>>>>
>>>>>>>> The following modules did not run:
>>>>>>>> Failed: Contrast Manager
>>>>>>>> ========================================================================
>>>>>>>>
>>>>>>>> When creating the contrast image using contrast manager after
>>>>>>>> selecting RESULTS, it works, BUT the result is surprising: the
>>>>>>>> main
>>>>>>>> effect of group is over the whole brain!
>>>>>>>>
>>>>>>>> Any idea what goes wrong?
>>>>>>>>
>>>>>>>> Thanks for your time.
>>>>>>>>
>>>>>>>> Cheers,
>>>>>>>>
>>>>>>>> Yann
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> <Attachment1.png><Attachment2.png>
>>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>> --
>>>> Guillaume Flandin, PhD
>>>> Wellcome Centre for Human Neuroimaging
>>>> UCL Queen Square Institute of Neurology
>>>> London WC1N 3BG
>>>
>>>
>>
>>
|