help > 2nd-level correlation coefficients
Showing 1-3 of 3 posts
Jan 14, 2022 05:01 AM | Suneel Banerjee
2nd-level correlation coefficients
Hi Alfonso,
I recently ran an analysis that included two sessions (task/rest) for each subjects and two 2nd level covariates that group the subjects based on the stimulus modality they received. After completing the analysis, I took the 'ROI.mat' file from the one of the contrasts in the secondlevel results directory.
I notice the 'h' field in this structure, which I understand to be the Fischer-transformed correlation coefficients. My question is, which connections do these correlations describe? Even though there are multiple possible contrasts comparing the sessions and the within-subjects effects, should there not be four sets of correlation values? How do correlation coefficients change across contrasts?
Any help would be much appreciated. Thank you!
I recently ran an analysis that included two sessions (task/rest) for each subjects and two 2nd level covariates that group the subjects based on the stimulus modality they received. After completing the analysis, I took the 'ROI.mat' file from the one of the contrasts in the secondlevel results directory.
I notice the 'h' field in this structure, which I understand to be the Fischer-transformed correlation coefficients. My question is, which connections do these correlations describe? Even though there are multiple possible contrasts comparing the sessions and the within-subjects effects, should there not be four sets of correlation values? How do correlation coefficients change across contrasts?
Any help would be much appreciated. Thank you!
Jan 17, 2022 06:01 PM | Alfonso Nieto-Castanon - Boston University
RE: 2nd-level correlation coefficients
Hi Suneel,
In general the ROI.h field represents the "contrast values", determined by a linear combination of the second-level GLM regressor coefficients as determined by your between-subjects contrast vector/matrix.
In particular, in your analysis ROI(i).h(j) represents the difference between the two groups (Visual - Auditory) in the difference between conditions (task - rest) in the connectivity between the i'th and j'th ROIs
i.e. ( connectivity@Task_Visual - connectivity@Rest_Visual ) - ( connectivity@Task_Auditory - connectivity@Rest_Auditory )
Best
Alfonso
Originally posted by Suneel Banerjee:
In general the ROI.h field represents the "contrast values", determined by a linear combination of the second-level GLM regressor coefficients as determined by your between-subjects contrast vector/matrix.
In particular, in your analysis ROI(i).h(j) represents the difference between the two groups (Visual - Auditory) in the difference between conditions (task - rest) in the connectivity between the i'th and j'th ROIs
i.e. ( connectivity@Task_Visual - connectivity@Rest_Visual ) - ( connectivity@Task_Auditory - connectivity@Rest_Auditory )
Best
Alfonso
Originally posted by Suneel Banerjee:
Hi Alfonso,
I recently ran an analysis that included two sessions (task/rest) for each subjects and two 2nd level covariates that group the subjects based on the stimulus modality they received. After completing the analysis, I took the 'ROI.mat' file from the one of the contrasts in the secondlevel results directory.
I notice the 'h' field in this structure, which I understand to be the Fischer-transformed correlation coefficients. My question is, which connections do these correlations describe? Even though there are multiple possible contrasts comparing the sessions and the within-subjects effects, should there not be four sets of correlation values? How do correlation coefficients change across contrasts?
Any help would be much appreciated. Thank you!
I recently ran an analysis that included two sessions (task/rest) for each subjects and two 2nd level covariates that group the subjects based on the stimulus modality they received. After completing the analysis, I took the 'ROI.mat' file from the one of the contrasts in the secondlevel results directory.
I notice the 'h' field in this structure, which I understand to be the Fischer-transformed correlation coefficients. My question is, which connections do these correlations describe? Even though there are multiple possible contrasts comparing the sessions and the within-subjects effects, should there not be four sets of correlation values? How do correlation coefficients change across contrasts?
Any help would be much appreciated. Thank you!
Jan 25, 2022 05:01 AM | Suneel Banerjee
RE: 2nd-level correlation coefficients
Thank you so much for the clarification! When looking at the ROI.h
field for in a project with just a single session and no
between-subjects contrasts, would this simply be the
Fischer-transformed correlation coefficients between the ROI time
series?