Dear CONN experts,
I am interested in examining the effect of an intervention on the connectivity of different ROIs. After running my 2nd-level 2×2 mixed ANOVA (Group × Time), I noticed important connectivity differences between groups at baseline. I expect those differences but is there a way to compute a 2nd-level ANCOVA (i.e., entering baseline connectivity as a covariate) to control for baseline differences in connectivity, either voxel-wise or at the ROI level?
Design:
2 groups
2 tasks: VID, CTRL (contrast : VID > CTRL)
2 sessions: PRE, POST
Current setup:
- Subject effects : [1 -1] (Group 1 > Group 2)
- Conditions : [1 -1 -1 1] (Post VID – Post CTRL) – (Pre VID – Pre CTRL)
I have 25 ROIs and I am interested in both seed-to-voxel and ROI-to-ROI connectivity.
Any guidance would be very helpful.
Thanks in
advance,
Gloria
Dear Gloria,
Sorry that ANCOVA design would require your second-level covariates to be voxel-specific (or connection-specific for RRC analyses) and that functionality is not available in CONN's second-level analyses (2nd-level covariates apply equally to all voxels and all connections evaluated).
Perhaps a reasonable work-around would be to perform a conjunction to exclude from your results those which show baseline differences between groups (i.e. look at the same 2x2 interaction analysis, but then exclude from these results any voxels or connections that show significant connectivity differences pre intervention).
Another approach would be to first identify what factor may be driving those pre-intervention differences (e.g. perhaps they reflect differences in some demographic, clinical, or behavioral measures between the groups?) and then perform the same interaction analyses but adding those explanatory covariates as a control 2nd-level covariate (e.g. changing in your original analysis the between-subjects contrast to [1 -1 0] with Group1, Group2, and BehaviorCovariate as subject effects)
Hope this helps
Alfonso
Originally posted by Gloria Leblond-B:
Dear CONN experts,
I am interested in examining the effect of an intervention on the connectivity of different ROIs. After running my 2nd-level 2×2 mixed ANOVA (Group × Time), I noticed important connectivity differences between groups at baseline. I expect those differences but is there a way to compute a 2nd-level ANCOVA (i.e., entering baseline connectivity as a covariate) to control for baseline differences in connectivity, either voxel-wise or at the ROI level?
Design:
2 groups
2 tasks: VID, CTRL (contrast : VID > CTRL)
2 sessions: PRE, POST
Current setup:
- Subject effects : [1 -1] (Group 1 > Group 2)
- Conditions : [1 -1 -1 1] (Post VID – Post CTRL) – (Pre VID – Pre CTRL)
I have 25 ROIs and I am interested in both seed-to-voxel and ROI-to-ROI connectivity.
Any guidance would be very helpful.
Thanks in advance,
Gloria