[color=#000000]Hi[/color]
[color=#000000]I do not know the details of how SPM creates these plots but my guess would be that you may be plotting the data from nearby but not exactly the same voxel locations. A more robust way to create these plots would be to first extract the connectivity values aggregated across all voxels within your cluster of interest, instead of selecting a single voxel (e.g. this can be done using the 'extract values' button in CONN's results explorer window), and then simply plotting those connectivity values against your behavioral covariate (e.g. this can be done using [i]Tools.Calculator [/i]in CONN) [/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by konstantina ki:[/i][quote]Hi Alfonso!
I am doing a resting-state functional connectivity analysis with a behavioural covariate (i.e., I want to see if my behavioural variable affects the RS seed-based connectivity amongst individuals). To visualise my second-level results, I click on SPM button in the CONN results display option.
To show the scatterplot of connectivity against the behavioural variable, I select plot->fitted responses --> predicted-->against variable and I select my behavioural contrast generated by CONN.
What I notice is that the scale of y axis (the connectivity values) can change from time to time. Although the pattern of the points seems to remain the same, it seems that some times y axis is expanded and sometimes is reduced. This can lead to few values becoming negative for example - despite the pattern of the points being similar.
This occurs even if I do not select 'recompute second level' but just 'load existing values'.
I cannot understand why this happens and what I can do to prevent SPM or CONN from doing this.
I assume that these are arbitrary units and so it shouldn't matter, right?
I need to be able to produce the same plots on every execution.
Can you please help?
Thanks,
Konstantina[/quote]
[color=#000000]I do not know the details of how SPM creates these plots but my guess would be that you may be plotting the data from nearby but not exactly the same voxel locations. A more robust way to create these plots would be to first extract the connectivity values aggregated across all voxels within your cluster of interest, instead of selecting a single voxel (e.g. this can be done using the 'extract values' button in CONN's results explorer window), and then simply plotting those connectivity values against your behavioral covariate (e.g. this can be done using [i]Tools.Calculator [/i]in CONN) [/color]
[color=#000000]Hope this helps[/color]
[color=#000000]Alfonso[/color]
[i]Originally posted by konstantina ki:[/i][quote]Hi Alfonso!
I am doing a resting-state functional connectivity analysis with a behavioural covariate (i.e., I want to see if my behavioural variable affects the RS seed-based connectivity amongst individuals). To visualise my second-level results, I click on SPM button in the CONN results display option.
To show the scatterplot of connectivity against the behavioural variable, I select plot->fitted responses --> predicted-->against variable and I select my behavioural contrast generated by CONN.
What I notice is that the scale of y axis (the connectivity values) can change from time to time. Although the pattern of the points seems to remain the same, it seems that some times y axis is expanded and sometimes is reduced. This can lead to few values becoming negative for example - despite the pattern of the points being similar.
This occurs even if I do not select 'recompute second level' but just 'load existing values'.
I cannot understand why this happens and what I can do to prevent SPM or CONN from doing this.
I assume that these are arbitrary units and so it shouldn't matter, right?
I need to be able to produce the same plots on every execution.
Can you please help?
Thanks,
Konstantina[/quote]