help
help > RE: Graph Theory correlation measures
Aug 28, 2015 12:08 AM | Alfonso Nieto-Castanon - Boston University
RE: Graph Theory correlation measures
Hi Ben,
Whether and edge passes the threshold or not is determined separately for each subject, not based on the joint/averaged effect across subjects. The reason is simply because we want to be able to compute subject-specific networks in order to then test differences in these network properties across subjects. To clarify, graphs/networks are computed separately for each subject (and for each condition if you have multiple conditions) by thresholding the subject/condition-specific ROI-to-ROI correlations. Then, again for each subject/condition separately, the corresponding network properties (e.g. global efficiency, local efficiency, etc.) are estimated, and last those measures are entered into your second-level model (e.g. to evaluate differences between groups or conditions).
The Z values that are being thresholded to compute the original networks are stored in conn*/results/firstlevel/ANALYSIS_01/resultsROI_Condition*.mat (each of this files contains a three-dimensional matrix Z containing the Fisher-transformed correlation values between each pair of ROIs and for each subject).
Hope this helps
Alfonso
Originally posted by Ben Crittenden:
Whether and edge passes the threshold or not is determined separately for each subject, not based on the joint/averaged effect across subjects. The reason is simply because we want to be able to compute subject-specific networks in order to then test differences in these network properties across subjects. To clarify, graphs/networks are computed separately for each subject (and for each condition if you have multiple conditions) by thresholding the subject/condition-specific ROI-to-ROI correlations. Then, again for each subject/condition separately, the corresponding network properties (e.g. global efficiency, local efficiency, etc.) are estimated, and last those measures are entered into your second-level model (e.g. to evaluate differences between groups or conditions).
The Z values that are being thresholded to compute the original networks are stored in conn*/results/firstlevel/ANALYSIS_01/resultsROI_Condition*.mat (each of this files contains a three-dimensional matrix Z containing the Fisher-transformed correlation values between each pair of ROIs and for each subject).
Hope this helps
Alfonso
Originally posted by Ben Crittenden:
Hello,
I thought I'd add to this thread as my question is quite similar...
I'd like to extract the ROI-ROI correlation values from a second level analysis, that correspond to what is used to determine the threshold in the graph-theory window of the gui.
Currently I can get the correlation values for a particular ROI by doing:
.h i.e. the effect size or i.e. averaging over the single-subject fisher Z values, which I believe should give the same result.
I thought that this would be the value that is used to determine whether an edge passes the threshold set by the correlation coefficient value in the graph theory window, but it doesn't appear to:
e.g. I have an edge between IPS and PFC with a correlation/effect size of 0.59, yet if I set the threshold up to 0.62 I still see the edge appear in the figure.
Is it possible to get the ROI-ROI correlation values that are compared against the correlation coefficient threshold that is set by the user in the graph theory window of the GUI? Or how is this computed?
I think the toolbox is great - many thanks!
Ben
I thought I'd add to this thread as my question is quite similar...
I'd like to extract the ROI-ROI correlation values from a second level analysis, that correspond to what is used to determine the threshold in the graph-theory window of the gui.
Currently I can get the correlation values for a particular ROI by doing:
.h i.e. the effect size or i.e. averaging over the single-subject fisher Z values, which I believe should give the same result.
I thought that this would be the value that is used to determine whether an edge passes the threshold set by the correlation coefficient value in the graph theory window, but it doesn't appear to:
e.g. I have an edge between IPS and PFC with a correlation/effect size of 0.59, yet if I set the threshold up to 0.62 I still see the edge appear in the figure.
Is it possible to get the ROI-ROI correlation values that are compared against the correlation coefficient threshold that is set by the user in the graph theory window of the GUI? Or how is this computed?
I think the toolbox is great - many thanks!
Ben
Threaded View
| Title | Author | Date |
|---|---|---|
| Aiden Arnold | Aug 1, 2012 | |
| Shukti Ramkiran | May 3, 2016 | |
| Alfonso Nieto-Castanon | Aug 2, 2012 | |
| Anna Zilverstand | Sep 17, 2015 | |
| Alfonso Nieto-Castanon | Sep 17, 2015 | |
| Anna Zilverstand | Sep 17, 2015 | |
| Alfonso Nieto-Castanon | Sep 17, 2015 | |
| Anna Zilverstand | Sep 21, 2015 | |
| Alfonso Nieto-Castanon | Sep 25, 2015 | |
| Anna Zilverstand | Sep 25, 2015 | |
| Alfonso Nieto-Castanon | Sep 25, 2015 | |
| Anna Zilverstand | Sep 28, 2015 | |
| Alfonso Nieto-Castanon | Sep 28, 2015 | |
| Anna Zilverstand | Sep 28, 2015 | |
| Alfonso Nieto-Castanon | Sep 29, 2015 | |
| Anna Zilverstand | Sep 29, 2015 | |
| Anna Zilverstand | Sep 25, 2015 | |
| Anna Zilverstand | Sep 21, 2015 | |
| Anna Zilverstand | Sep 17, 2015 | |
| Ben Crittenden | Aug 27, 2015 | |
| Alfonso Nieto-Castanon | Aug 28, 2015 | |
| Ben Crittenden | Sep 18, 2015 | |
| Ben Crittenden | Sep 1, 2015 | |
| Aiden Arnold | Aug 3, 2012 | |
| Alfonso Nieto-Castanon | Aug 13, 2012 | |