help
help > RE: Possibly different connectivty values
Sep 9, 2015 05:09 AM | Alfonso Nieto-Castanon - Boston University
RE: Possibly different connectivty values
Dear Pravesh,
I looked at the ROI-to-ROI stats for the 'Free_Attention' condition in both analyses and I do not quite see the bias that you mentioned (see attached figure). There are some small differences, but I believe those differences are possibly due to the SZ_only experiment having one additional task condition ('Hallucination Attention') that was not present in the combined analyses (the inclusion of this additional condition may produce subtle differences in the two analyses through differences in: a) the computation of White/CSF principal components -which remove task and first-level covariates prior to PCA decomposition in noise ROIs-; and b) the Denoising step -which removes main BOLD activation effects associated with the presence of this condition from the BOLD signal before computing correlation measures-. If you could perhaps try adding that same condition to to the combined analyses -or removing that condition from the SZ_only analyses- and repeating the analysis steps I believe the two pipelines should result in exactly the same values (as far as I can tell there were no other differences between the two experiments)
Hope this helps
Alfonso
Originally posted by Pravesh Parekh:
I looked at the ROI-to-ROI stats for the 'Free_Attention' condition in both analyses and I do not quite see the bias that you mentioned (see attached figure). There are some small differences, but I believe those differences are possibly due to the SZ_only experiment having one additional task condition ('Hallucination Attention') that was not present in the combined analyses (the inclusion of this additional condition may produce subtle differences in the two analyses through differences in: a) the computation of White/CSF principal components -which remove task and first-level covariates prior to PCA decomposition in noise ROIs-; and b) the Denoising step -which removes main BOLD activation effects associated with the presence of this condition from the BOLD signal before computing correlation measures-. If you could perhaps try adding that same condition to to the combined analyses -or removing that condition from the SZ_only analyses- and repeating the analysis steps I believe the two pipelines should result in exactly the same values (as far as I can tell there were no other differences between the two experiments)
Hope this helps
Alfonso
Originally posted by Pravesh Parekh:
Dear Dr.
Alfonso,
Please find attached a zip file containing both the project files and the second level analysis folder contained inside each of them. Please let me know if these are the files that you were referring to in your response.
Please do let us know if the differences are because of some error that might have occurred on our end!
Regards
Pravesh Parekh
Originally posted by Alfonso Nieto-Castanon:
Please find attached a zip file containing both the project files and the second level analysis folder contained inside each of them. Please let me know if these are the files that you were referring to in your response.
Please do let us know if the differences are because of some error that might have occurred on our end!
Regards
Pravesh Parekh
Originally posted by Alfonso Nieto-Castanon:
Dear
Pravesh,
You are right that, from your description, you should be seeing exactly the same results in those two analyses. If you could please send me the two ROI.mat files (one associated with each ROI-to-ROI analysis), and perhaps also the two conn*.mat files (one associated with each conn project), perhaps that could help me figure out what these differences may reflect.
Best
Alfonso
Originally posted by Pravesh Parekh:
You are right that, from your description, you should be seeing exactly the same results in those two analyses. If you could please send me the two ROI.mat files (one associated with each ROI-to-ROI analysis), and perhaps also the two conn*.mat files (one associated with each conn project), perhaps that could help me figure out what these differences may reflect.
Best
Alfonso
Originally posted by Pravesh Parekh:
Dear Dr. Alfonso,
We have two groups of subjects: healthy (HS) and patients with schizophrenia (SZ). The SZ group can be sub-divided into patients with hallucinations (AVH+) and those who do not have hallucinations (AVH-). We ran two set of analysis (independent conn projects): the first had
HS and SZ groups (included as second level covariates). In the second project, we only included SZ subjects and had two second level covariates (AVH+ and AVH-). Both projects were analysed with the same parameters till second level.
For Project1 (HS and SZ groups):
- Selected only SZ group (in the subject effects list)
- Clicked on Results Explorer
- Selected all ROIs (and de-selected F test)
For Project2 (only SZ group; AVH+ and AVH-)
- Selected All Subjects in the subjects effects list
- Clicked on Results Explorer
- Selected all ROIs (and de-selected F test)
Upon comparing both the outputs, we found that:
- Overall connectivity pattern is almost the same (there are some minor differences)
- There is a difference in the T statistic value that we see (all values are shifted)
How do we account for the difference in the T values (as it is essentially the same set of subjects that are being analysed)? Shouldn't the values be same or are we doing something wrong somewhere?
The output from both the runs are also attached. We have ensured that in both the runs the same condition was selected (and we in fact did run the analysis a second time just to be sure).
Regards,
Pravesh Parekh
We have two groups of subjects: healthy (HS) and patients with schizophrenia (SZ). The SZ group can be sub-divided into patients with hallucinations (AVH+) and those who do not have hallucinations (AVH-). We ran two set of analysis (independent conn projects): the first had
HS and SZ groups (included as second level covariates). In the second project, we only included SZ subjects and had two second level covariates (AVH+ and AVH-). Both projects were analysed with the same parameters till second level.
For Project1 (HS and SZ groups):
- Selected only SZ group (in the subject effects list)
- Clicked on Results Explorer
- Selected all ROIs (and de-selected F test)
For Project2 (only SZ group; AVH+ and AVH-)
- Selected All Subjects in the subjects effects list
- Clicked on Results Explorer
- Selected all ROIs (and de-selected F test)
Upon comparing both the outputs, we found that:
- Overall connectivity pattern is almost the same (there are some minor differences)
- There is a difference in the T statistic value that we see (all values are shifted)
How do we account for the difference in the T values (as it is essentially the same set of subjects that are being analysed)? Shouldn't the values be same or are we doing something wrong somewhere?
The output from both the runs are also attached. We have ensured that in both the runs the same condition was selected (and we in fact did run the analysis a second time just to be sure).
Regards,
Pravesh Parekh
Threaded View
| Title | Author | Date |
|---|---|---|
| Pravesh Parekh | Sep 1, 2015 | |
| Alfonso Nieto-Castanon | Sep 3, 2015 | |
| Pravesh Parekh | Sep 4, 2015 | |
| Alfonso Nieto-Castanon | Sep 9, 2015 | |
| Pravesh Parekh | Sep 1, 2015 | |