open-discussion > group comparison
Showing 1-3 of 3 posts
Jul 16, 2015 11:07 AM | Lars Michels
group comparison
Hi all
I have two groups and i want to compare them with respect to connectivity differences.
How can I be sure about the direction of correlations in the difference image?
For example, if I code 1 -1, and I see only blue lines. Does this just mean that correlations are stronger for group 2 (irrespective of being positive or anticorrelations)?
In constrast, If I change to one-sided (positive), does this only show POSITIVE correlations for the same contrast, i.e. 1 -1, or does this only belongs to the group assignment? I think the latter is the case, because when I estimated -1 1, I get the same connections (shown as blue lines) as for the reversed contrast (red lines) if turn to from one-sided (positive) to one-sided (negative)
it is difficult to intepret group differences if not group assignemt can have 2 signs but also correlation
Any ideas how to just visualize group differences for regions showing POSITIVE correlations?
Best wishes and Thanks
Lars
I have two groups and i want to compare them with respect to connectivity differences.
How can I be sure about the direction of correlations in the difference image?
For example, if I code 1 -1, and I see only blue lines. Does this just mean that correlations are stronger for group 2 (irrespective of being positive or anticorrelations)?
In constrast, If I change to one-sided (positive), does this only show POSITIVE correlations for the same contrast, i.e. 1 -1, or does this only belongs to the group assignment? I think the latter is the case, because when I estimated -1 1, I get the same connections (shown as blue lines) as for the reversed contrast (red lines) if turn to from one-sided (positive) to one-sided (negative)
it is difficult to intepret group differences if not group assignemt can have 2 signs but also correlation
Any ideas how to just visualize group differences for regions showing POSITIVE correlations?
Best wishes and Thanks
Lars
Jul 16, 2015 09:07 PM | Alfonso Nieto-Castanon - Boston University
RE: group comparison
Hi Lars,
When comparing two groups (e.g. selecting GroupA and GroupB and entering a 1 -1 contrast) the color-coding represents the sign/direction of the selected between-group differences in connectivity (e.g. in the example above red/positive results correspond to GroupA>GroupB and blue/negative results correspond to GroupB>GroupA). You are right that these differences are independent of the sign of the actual within-group correlations (e.g. a blue/negative group difference -GroupB>GroupA- only means that the correlations in GroupB are higher -i.e. more positive or less negative- than the correlations in GroupA, but it does not tell you anything about the sign of those within-group correlations).
To better interpret those between-group differences I would suggest to first look at your results in the main CONN gui (second-level results ROI-to-ROI tab). Select there your seed region, the two groups GroupA and GroupB and enter the 1 -1 contrast. Then you can click in any target ROI (either in the results table or in the results brain-display) and that will show you a small popup window with a bar-graph containing the within-group correlations (mean and standard errors of the connectivity/correlation within groupA and within groupB separately), which will allow you to better grasp simultaneously both the sign of the individual within-group correlations as well as the directionality of the between-group differences.
Hope this helps
Alfonso
Originally posted by Lars Michels:
When comparing two groups (e.g. selecting GroupA and GroupB and entering a 1 -1 contrast) the color-coding represents the sign/direction of the selected between-group differences in connectivity (e.g. in the example above red/positive results correspond to GroupA>GroupB and blue/negative results correspond to GroupB>GroupA). You are right that these differences are independent of the sign of the actual within-group correlations (e.g. a blue/negative group difference -GroupB>GroupA- only means that the correlations in GroupB are higher -i.e. more positive or less negative- than the correlations in GroupA, but it does not tell you anything about the sign of those within-group correlations).
To better interpret those between-group differences I would suggest to first look at your results in the main CONN gui (second-level results ROI-to-ROI tab). Select there your seed region, the two groups GroupA and GroupB and enter the 1 -1 contrast. Then you can click in any target ROI (either in the results table or in the results brain-display) and that will show you a small popup window with a bar-graph containing the within-group correlations (mean and standard errors of the connectivity/correlation within groupA and within groupB separately), which will allow you to better grasp simultaneously both the sign of the individual within-group correlations as well as the directionality of the between-group differences.
Hope this helps
Alfonso
Originally posted by Lars Michels:
Hi all
I have two groups and i want to compare them with respect to connectivity differences.
How can I be sure about the direction of correlations in the difference image?
For example, if I code 1 -1, and I see only blue lines. Does this just mean that correlations are stronger for group 2 (irrespective of being positive or anticorrelations)?
In constrast, If I change to one-sided (positive), does this only show POSITIVE correlations for the same contrast, i.e. 1 -1, or does this only belongs to the group assignment? I think the latter is the case, because when I estimated -1 1, I get the same connections (shown as blue lines) as for the reversed contrast (red lines) if turn to from one-sided (positive) to one-sided (negative)
it is difficult to intepret group differences if not group assignemt can have 2 signs but also correlation
Any ideas how to just visualize group differences for regions showing POSITIVE correlations?
Best wishes and Thanks
Lars
I have two groups and i want to compare them with respect to connectivity differences.
How can I be sure about the direction of correlations in the difference image?
For example, if I code 1 -1, and I see only blue lines. Does this just mean that correlations are stronger for group 2 (irrespective of being positive or anticorrelations)?
In constrast, If I change to one-sided (positive), does this only show POSITIVE correlations for the same contrast, i.e. 1 -1, or does this only belongs to the group assignment? I think the latter is the case, because when I estimated -1 1, I get the same connections (shown as blue lines) as for the reversed contrast (red lines) if turn to from one-sided (positive) to one-sided (negative)
it is difficult to intepret group differences if not group assignemt can have 2 signs but also correlation
Any ideas how to just visualize group differences for regions showing POSITIVE correlations?
Best wishes and Thanks
Lars
May 24, 2021 10:05 PM | Aishvarya Shri Rajasimman - AIIMS, Bhubaneswar
RE: group comparison
Originally posted by Alfonso Nieto-Castanon:
During second level analysis between cases and controls,
keeping cases > controls or controls> cases, in a particular seed - I am observing the connection appears blue in both the scenarios.
however, the T score is positive in cases >control (ie., 1 -1) setting and negative with (-1 1) setting.
So can I infer that the blue color here represents the increased connections among cases compared to controls? I am confused. Have attached a screenshot as well. Please guide me
Thanks in advance.
Regards,
Aishvarya
Hi Lars,
When comparing two groups (e.g. selecting GroupA and GroupB and entering a 1 -1 contrast) the color-coding represents the sign/direction of the selected between-group differences in connectivity (e.g. in the example above red/positive results correspond to GroupA>GroupB and blue/negative results correspond to GroupB>GroupA). You are right that these differences are independent of the sign of the actual within-group correlations (e.g. a blue/negative group difference -GroupB>GroupA- only means that the correlations in GroupB are higher -i.e. more positive or less negative- than the correlations in GroupA, but it does not tell you anything about the sign of those within-group correlations).
To better interpret those between-group differences I would suggest to first look at your results in the main CONN gui (second-level results ROI-to-ROI tab). Select there your seed region, the two groups GroupA and GroupB and enter the 1 -1 contrast. Then you can click in any target ROI (either in the results table or in the results brain-display) and that will show you a small popup window with a bar-graph containing the within-group correlations (mean and standard errors of the connectivity/correlation within groupA and within groupB separately), which will allow you to better grasp simultaneously both the sign of the individual within-group correlations as well as the directionality of the between-group differences.
Hope this helps
Alfonso
Dear friends,When comparing two groups (e.g. selecting GroupA and GroupB and entering a 1 -1 contrast) the color-coding represents the sign/direction of the selected between-group differences in connectivity (e.g. in the example above red/positive results correspond to GroupA>GroupB and blue/negative results correspond to GroupB>GroupA). You are right that these differences are independent of the sign of the actual within-group correlations (e.g. a blue/negative group difference -GroupB>GroupA- only means that the correlations in GroupB are higher -i.e. more positive or less negative- than the correlations in GroupA, but it does not tell you anything about the sign of those within-group correlations).
To better interpret those between-group differences I would suggest to first look at your results in the main CONN gui (second-level results ROI-to-ROI tab). Select there your seed region, the two groups GroupA and GroupB and enter the 1 -1 contrast. Then you can click in any target ROI (either in the results table or in the results brain-display) and that will show you a small popup window with a bar-graph containing the within-group correlations (mean and standard errors of the connectivity/correlation within groupA and within groupB separately), which will allow you to better grasp simultaneously both the sign of the individual within-group correlations as well as the directionality of the between-group differences.
Hope this helps
Alfonso
During second level analysis between cases and controls,
keeping cases > controls or controls> cases, in a particular seed - I am observing the connection appears blue in both the scenarios.
however, the T score is positive in cases >control (ie., 1 -1) setting and negative with (-1 1) setting.
So can I infer that the blue color here represents the increased connections among cases compared to controls? I am confused. Have attached a screenshot as well. Please guide me
Thanks in advance.
Regards,
Aishvarya