help > how to run second level analysis with batch?
Showing 1-4 of 4 posts
May 17, 2017 07:05 PM | ml90
how to run second level analysis with batch?
Hi
I saved a conn project (batch) that included all the standard steps. Now I would like to run second level analysis so I defined
clear BATCH
BATCH.filename = 'conn_project.mat'
BATCH.foldername = 'New'
BATCH.Results.done = 1
BATCH.Results.between_conditions.effect_names = {'A','B'}
BATCH.Results.between_conditions.contrast = [-1 1]
conn_batch(BATCH)
But nothing happens. Is there something missing from the code above?
Thanks in advance.
Michael
I saved a conn project (batch) that included all the standard steps. Now I would like to run second level analysis so I defined
clear BATCH
BATCH.filename = 'conn_project.mat'
BATCH.foldername = 'New'
BATCH.Results.done = 1
BATCH.Results.between_conditions.effect_names = {'A','B'}
BATCH.Results.between_conditions.contrast = [-1 1]
conn_batch(BATCH)
But nothing happens. Is there something missing from the code above?
Thanks in advance.
Michael
May 18, 2017 03:05 AM | Alfonso Nieto-Castanon - Boston University
RE: how to run second level analysis with batch?
Hi,
The issue here seems to be that the BATCH structure defined is missing the BATCH.Results.between_subjects values (specifying your between-subject effects/contrasts). See help conn_batch for additional details
Best
Alfonso
Originally posted by ml90:
The issue here seems to be that the BATCH structure defined is missing the BATCH.Results.between_subjects values (specifying your between-subject effects/contrasts). See help conn_batch for additional details
Best
Alfonso
Originally posted by ml90:
Hi
I saved a conn project (batch) that included all the standard steps. Now I would like to run second level analysis so I defined
clear BATCH
BATCH.filename = 'conn_project.mat'
BATCH.foldername = 'New'
BATCH.Results.done = 1
BATCH.Results.between_conditions.effect_names = {'A','B'}
BATCH.Results.between_conditions.contrast = [-1 1]
conn_batch(BATCH)
But nothing happens. Is there something missing from the code above?
Thanks in advance.
Michael
I saved a conn project (batch) that included all the standard steps. Now I would like to run second level analysis so I defined
clear BATCH
BATCH.filename = 'conn_project.mat'
BATCH.foldername = 'New'
BATCH.Results.done = 1
BATCH.Results.between_conditions.effect_names = {'A','B'}
BATCH.Results.between_conditions.contrast = [-1 1]
conn_batch(BATCH)
But nothing happens. Is there something missing from the code above?
Thanks in advance.
Michael
May 18, 2017 09:05 PM | ml90
RE: how to run second level analysis with batch?
Hi Alfonso,
Yes that seemed to have been the problem, thank you.
However I'm having a problem now when I try to display second level results in SPM. I tried to run a couple of second level analysis using the conn gui. When there are significant voxels, and I load the resulting SPM.mat in SPM, it works fine.
However when there are no significant voxels, and I try to load the results in SPM, it takes a very long time and displays this warning message a few times on the matlab window
Warning: Returning NaN for out of range arguments
> In spm_Pcdf at 90
In spm_P_RF at 112
In spm_P at 39
In spm_list at 503
In spm_list at 136
The place where the glass brain is supposed to appear is completely black, and the statistics show a T of over 300, and a significant cluster of over 60000 voxels.
I also noticed that the mask.nii file stored in the second level analysis folder contains only ones (there are no zeros).
Best, Michael
Yes that seemed to have been the problem, thank you.
However I'm having a problem now when I try to display second level results in SPM. I tried to run a couple of second level analysis using the conn gui. When there are significant voxels, and I load the resulting SPM.mat in SPM, it works fine.
However when there are no significant voxels, and I try to load the results in SPM, it takes a very long time and displays this warning message a few times on the matlab window
Warning: Returning NaN for out of range arguments
> In spm_Pcdf at 90
In spm_P_RF at 112
In spm_P at 39
In spm_list at 503
In spm_list at 136
The place where the glass brain is supposed to appear is completely black, and the statistics show a T of over 300, and a significant cluster of over 60000 voxels.
I also noticed that the mask.nii file stored in the second level analysis folder contains only ones (there are no zeros).
Best, Michael
May 22, 2017 07:05 PM | Alfonso Nieto-Castanon - Boston University
RE: how to run second level analysis with batch?
Hi Michael,
Those warning message and strange-looking results are typically indicative of either a not-terribly-specific null model or (perhaps more likely) an incorrectly-defined second-level model. In particular the out-of-range message indicates that you are seeing extremely large T-values in your stats (which is also corroborated by the T>300 values) so these are not results with no significant voxels but rather results where every voxel is significant. I would check the second-level model that you are specifying, are you seeing in the second-level results CONN gui a red button that reads 'WARNING!" where you would usually see a "n=#" message? (if so that indicates the latter explanation; in this case simply click on that button to see more details about why CONN believes your model may be incorrectly defined). If this is not the case please let me know more details about the specific second-level model that you are defining.
Hope this helps
Alfonso
Originally posted by ml90:
Those warning message and strange-looking results are typically indicative of either a not-terribly-specific null model or (perhaps more likely) an incorrectly-defined second-level model. In particular the out-of-range message indicates that you are seeing extremely large T-values in your stats (which is also corroborated by the T>300 values) so these are not results with no significant voxels but rather results where every voxel is significant. I would check the second-level model that you are specifying, are you seeing in the second-level results CONN gui a red button that reads 'WARNING!" where you would usually see a "n=#" message? (if so that indicates the latter explanation; in this case simply click on that button to see more details about why CONN believes your model may be incorrectly defined). If this is not the case please let me know more details about the specific second-level model that you are defining.
Hope this helps
Alfonso
Originally posted by ml90:
Hi Alfonso,
Yes that seemed to have been the problem, thank you.
However I'm having a problem now when I try to display second level results in SPM. I tried to run a couple of second level analysis using the conn gui. When there are significant voxels, and I load the resulting SPM.mat in SPM, it works fine.
However when there are no significant voxels, and I try to load the results in SPM, it takes a very long time and displays this warning message a few times on the matlab window
Warning: Returning NaN for out of range arguments
> In spm_Pcdf at 90
In spm_P_RF at 112
In spm_P at 39
In spm_list at 503
In spm_list at 136
The place where the glass brain is supposed to appear is completely black, and the statistics show a T of over 300, and a significant cluster of over 60000 voxels.
I also noticed that the mask.nii file stored in the second level analysis folder contains only ones (there are no zeros).
Best, Michael
Yes that seemed to have been the problem, thank you.
However I'm having a problem now when I try to display second level results in SPM. I tried to run a couple of second level analysis using the conn gui. When there are significant voxels, and I load the resulting SPM.mat in SPM, it works fine.
However when there are no significant voxels, and I try to load the results in SPM, it takes a very long time and displays this warning message a few times on the matlab window
Warning: Returning NaN for out of range arguments
> In spm_Pcdf at 90
In spm_P_RF at 112
In spm_P at 39
In spm_list at 503
In spm_list at 136
The place where the glass brain is supposed to appear is completely black, and the statistics show a T of over 300, and a significant cluster of over 60000 voxels.
I also noticed that the mask.nii file stored in the second level analysis folder contains only ones (there are no zeros).
Best, Michael