Back to PLS HelpLV images - how does group affect them?
Posted on 10/22/10 12:09:07
Number of posts: 7
Hi there, sorry - yet another question!
I have a three group design, with 6 behavioural measures (using structural PLS and only one condition), I wonder if you could shed some light on how group affects the LV images
Am I right in presuming that, for a given LV, the grey matter system elucidated represents those voxels where there is a group difference in the correlation between GM density and behaviour? Or is that system simply those voxels that optimally correlate with behaviour across all groups, you then look at how the groups load on the LV to work out the involvement of group? I see the advantage of loading the groups in separately as any GM difference associated with group and not behaviour is factored out, but I'm unsure how group is involved in producing the LV.
Any help would be much appreciated! Thank you.
Anna
I have a three group design, with 6 behavioural measures (using structural PLS and only one condition), I wonder if you could shed some light on how group affects the LV images
Am I right in presuming that, for a given LV, the grey matter system elucidated represents those voxels where there is a group difference in the correlation between GM density and behaviour? Or is that system simply those voxels that optimally correlate with behaviour across all groups, you then look at how the groups load on the LV to work out the involvement of group? I see the advantage of loading the groups in separately as any GM difference associated with group and not behaviour is factored out, but I'm unsure how group is involved in producing the LV.
Any help would be much appreciated! Thank you.
Anna
Replies:
Untitled Post
I'm Online
Posted on 10/26/10 10:22:36
Number of posts: 229
Anna..
In the behavPLS, we do not remove main effects of group, as is done in the taskPLS.
So your LVs may reflect
1) regions with common GM-behav correlations across groups
2) regions with opposite GM-behav correlations
3) some mixture of the above..
You can get a better sense of what the LV is showing by focusing on the correlation overview, using the confidence intervals for the correlations to guide you as to which GM-behav correlations are driving the LV. We generally focus on those correlations where the correlation CI does not include zero
nancy
In the behavPLS, we do not remove main effects of group, as is done in the taskPLS.
So your LVs may reflect
1) regions with common GM-behav correlations across groups
2) regions with opposite GM-behav correlations
3) some mixture of the above..
You can get a better sense of what the LV is showing by focusing on the correlation overview, using the confidence intervals for the correlations to guide you as to which GM-behav correlations are driving the LV. We generally focus on those correlations where the correlation CI does not include zero
nancy
Login to reply to this topic.
