Dear experts,
I have completed a seed-to-voxel analysis and now I would like to extract the individual correlations between my seed and the cluster identified. How can I achieve? Do I need to use a program such as Marsbar?
Regards
Andrew

Dear Andrew,

The simplest way to do this is, from the 'seed to voxel results explorer', select 'export mask', then define a name for the mask file containing all suprathreshold voxels, and last, when prompted for additional options select 'Explore mask results'. This will launch rex which will extract the average individual correlations across each cluster identified and it will allow you to display those values and perform some simple statistics on them (it will also save a file named REX.mat which will contain among other things the average Z-scores across each individual cluster for each subject; see field params.ROIdata within the REX.mat file)

Hope this helps
Alfonso

Originally posted by Andrew Martin:

Dear Andrew,

To make this common process simpler, in the latest version of the toolbox (CONN14c) you may simply click on the 'import values' button in the seed-to-voxel results explorer. This will import the average connectivity values for each subject within each cluster as new 'second-level covariates'. You may then use Tools-> Calculator to view these values or to perform any additional post-hoc analysis of these values (and of course you may also use these as additional between-subject effects for other post-hoc seed-to-voxel or ROI-to-ROI analyses).

Hope this helps
Alfonso
Originally posted by Alfonso Nieto-Castanon:

Dear Alfonso,

I have a short question about your description to extract those values using the "import values" button. Are the resulting values z-scores?

Thank you for your help!

Best
Max

Originally posted by Alfonso Nieto-Castanon:

Hi Max,

The 'import values' results (stored as one or several new second-level covariates) contain average "connectivity" values within each cluster. The interpretation of these "connectivity" values depends on your choice of associated first-level analysis. For example, if you are analyzing seed-to-voxel maps and these 1st-level SBC analyses using bivariate correlation measures (the default settings) then the values at a voxel will be Fisher-transformed correlation coefficients between the selected seed and this voxel. Those values at each voxel are the ones that get aggregated across all voxels within each cluster and separately for each subject and exported as a new second-level covariate during the 'import values' step.

Hope this helps
Alfonso
Originally posted by max345:

 Dear all,

I have a short follow-up question regarding the import values tool.

I am studying a population with three main subgroups, let's say A, B, C. I have run a seed-based analysis and performed a t-test to compare two groups, for instance A and B, and found some significant clusters. If I later run 'import values' would I also obtain the connectivity from that seed to each significant cluster for the leftover group C?

I count on your help!

Best,
Simona

Hi all,

I have extracted connectivity values using the "import values" option for each individual in my analysis and using the calculator compare 2 groups while controlling for age and IQ (using contrast 1 -1 0 0). According to the results of the analysis in the calculator, the difference between groups are significant. However, when I export these same values from the "2nd-level covariates" list and enter them into a statistical software such as SPSS, the difference is no longer significant (even when including the same covariates). Why would this be? Is there some additional analysis CONN is running with the calculator that is not run in a standard ANOVA or linear regression model?

Thanks for any help,
Alex

Hi Alex

That's curious, the results should match exactly, in particular defining in CONN an analysis using the 2nd-level regressors [GroupA, GroupB, age, IQ] (assuming groupA and groupB are defined as 0/1 variables spanning all of your study subjects), with a between-subjects contrast [-1 1 0 0], should match exactly the results of an standard ANCOVA with age/IQ as covariates (assuming equal variances across groups). To figure out what might be the issue here, could you please check if the degrees of freedom reported in both analyses match?

Best
Alfonso
Originally posted by Alexandra Muratore: