help > Graph Theory Results Interpretation
Showing 1-5 of 5 posts
Apr 19, 2022 12:04 PM | fhaase
Graph Theory Results Interpretation
Hello everybody,
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
Apr 25, 2022 02:04 PM | fhaase
RE: Graph Theory Results Interpretation
So sorry to bother you again, but I did not make any progress
yet.
If anyone has an idea, please let me know. That would be really helpful!
Best,
Franziska
If anyone has an idea, please let me know. That would be really helpful!
Best,
Franziska
Apr 27, 2022 11:04 PM | Alfonso Nieto-Castanon - Boston University
RE: Graph Theory Results Interpretation
Hi Franziska,
Your [1 0] between-subjects contrast is computing the average pre-post change in global efficiency at the zero level of your age covariate. One possible guess here is that if the age covariate has not been centered then the zero-level of that covariate may not be terribly meaningful (and it could potentially cause the flip in sign that you are observing between the average global efficiency change in your sample and the expected global efficiency change extrapolated at age = 0). If that is the case then simply centering your age covariate (or defining a contrast [1 X] where X is the actual average value of the age covariate in your sample) will allow you to test the change in global efficiency while controlling for age effects (estimated change in the population at an age equal to the average age of your sample).
Hope this helps
Alfonso
Originally posted by fhaase:
Your [1 0] between-subjects contrast is computing the average pre-post change in global efficiency at the zero level of your age covariate. One possible guess here is that if the age covariate has not been centered then the zero-level of that covariate may not be terribly meaningful (and it could potentially cause the flip in sign that you are observing between the average global efficiency change in your sample and the expected global efficiency change extrapolated at age = 0). If that is the case then simply centering your age covariate (or defining a contrast [1 X] where X is the actual average value of the age covariate in your sample) will allow you to test the change in global efficiency while controlling for age effects (estimated change in the population at an age equal to the average age of your sample).
Hope this helps
Alfonso
Originally posted by fhaase:
Hello everybody,
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
May 2, 2022 12:05 PM | fhaase
RE: Graph Theory Results Interpretation
Hi Alfonso,
thank you so much for your reply and clarification! I mean-centered my age covariable as you suggested and there were no significant results; neither for graph theory nor for the ROI-to-ROI analysis which before resulted in a significant cluster in the salience network. Therefore, I now have another question regarding the ROI-to-ROI analysis. When using age without mean centering, there were significant results suggesting higher connectivity in the post test (see attachment page 1). As these were not there when I mean-centered age, I tested the effect of age in the ROI-to-ROI analysis (between-subjects contrast [0 1]), which resulted in the exact same significant cluster, but indicating higher FC values at pre with age (see attachment page 2). Am I correct in interpreting that age might be the limiting factor in this analysis leading to the significant results? The age range is between 18 and 21, except for two participants, who are 28 and therefore significantly older. When I exclude those participants, there are no significant results, too.
In addition to that, I am also having problems in interpreting the results from my seed based analysis (between condition contrast [1 -1]; seeds/sources: DMN and SN, between-sources contrast: any effects). The non-parametric permutation/randomization analysis leads to a very high number of significant clusters (41 for all subjects + age; 2067 for all subjects+ mean centered age; see attachment pages 4 and 5). As you can see, the clusters are very small (only 1 to 6) voxels, but have a very high mass, which is why they are declared significant. How should I treat those results?
Again, thank you for helping! I really appreciate the time and effort you put into answering all of our questions on the forum!
Best,
Franziska
PS: I had to create a PDF as I wasn't able to upload more than file, sorry!
Originally posted by Alfonso Nieto-Castanon:
thank you so much for your reply and clarification! I mean-centered my age covariable as you suggested and there were no significant results; neither for graph theory nor for the ROI-to-ROI analysis which before resulted in a significant cluster in the salience network. Therefore, I now have another question regarding the ROI-to-ROI analysis. When using age without mean centering, there were significant results suggesting higher connectivity in the post test (see attachment page 1). As these were not there when I mean-centered age, I tested the effect of age in the ROI-to-ROI analysis (between-subjects contrast [0 1]), which resulted in the exact same significant cluster, but indicating higher FC values at pre with age (see attachment page 2). Am I correct in interpreting that age might be the limiting factor in this analysis leading to the significant results? The age range is between 18 and 21, except for two participants, who are 28 and therefore significantly older. When I exclude those participants, there are no significant results, too.
In addition to that, I am also having problems in interpreting the results from my seed based analysis (between condition contrast [1 -1]; seeds/sources: DMN and SN, between-sources contrast: any effects). The non-parametric permutation/randomization analysis leads to a very high number of significant clusters (41 for all subjects + age; 2067 for all subjects+ mean centered age; see attachment pages 4 and 5). As you can see, the clusters are very small (only 1 to 6) voxels, but have a very high mass, which is why they are declared significant. How should I treat those results?
Again, thank you for helping! I really appreciate the time and effort you put into answering all of our questions on the forum!
Best,
Franziska
PS: I had to create a PDF as I wasn't able to upload more than file, sorry!
Originally posted by Alfonso Nieto-Castanon:
Hi
Franziska,
Your [1 0] between-subjects contrast is computing the average pre-post change in global efficiency at the zero level of your age covariate. One possible guess here is that if the age covariate has not been centered then the zero-level of that covariate may not be terribly meaningful (and it could potentially cause the flip in sign that you are observing between the average global efficiency change in your sample and the expected global efficiency change extrapolated at age = 0). If that is the case then simply centering your age covariate (or defining a contrast [1 X] where X is the actual average value of the age covariate in your sample) will allow you to test the change in global efficiency while controlling for age effects (estimated change in the population at an age equal to the average age of your sample).
Hope this helps
Alfonso
Originally posted by fhaase:
Your [1 0] between-subjects contrast is computing the average pre-post change in global efficiency at the zero level of your age covariate. One possible guess here is that if the age covariate has not been centered then the zero-level of that covariate may not be terribly meaningful (and it could potentially cause the flip in sign that you are observing between the average global efficiency change in your sample and the expected global efficiency change extrapolated at age = 0). If that is the case then simply centering your age covariate (or defining a contrast [1 X] where X is the actual average value of the age covariate in your sample) will allow you to test the change in global efficiency while controlling for age effects (estimated change in the population at an age equal to the average age of your sample).
Hope this helps
Alfonso
Originally posted by fhaase:
Hello everybody,
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
I am recently running into problems regarding the interpretation of the graph theroy results. Some information on the project:
I have 14 subjects in a longitudinal design with pre and post-measurements and I want to look at potential functional connectivity changes in the DMN and salience network (SN) ROIs between those measurements. Therefore, I am using as between-subject contrast [1 0] with age as control variable. The between-condition contrast is [1 -1] to see if there are changes from pre to post. In the graph theory results GUI I selected the DMN and SN ROIs and added the precuneus for the DMN. I applied a two-sided 0.15 cost threshold for the network edges and a two-sided p-FDR of 0.05 as analysis threshold.
I get some significant results for the right RPFC of the SN for global efficiency, cost and degree (see attachment for global efficiency results). As the node is displayed in blue. I am assuming that (because of the between-condition contrast) the global efficiency in that node is higher in the post-measurement. My issue is that I cannot replicate those results when exporting the data into a statistics program like SPSS. More specifically I conducted the means for the pre and the post measurement separatly from the CSV file and when comparing them it seems that the pre-measurement has higher values than the post-measurement although I expected it the other way around.
Am I understanding something wrong or how should I interpret the results from the GUI?
Any help is much appreciated!
Best,
Franziska
May 17, 2022 07:05 AM | fhaase
RE: Graph Theory Results Interpretation
Just checking if someone has an idea on this as I still was not
able to find a solution.
Thanks,
Franziska
Thanks,
Franziska