[Marsbar-users] Mean voxel value of SPM clusters - correlations HELP

[Julie C. B. <j1...@st...>]

Bonjour toute la monde,

I am using SPM5 for VBM analysis of structural MRI grey matter  
density.  I have done a correlation with a creativity measure and  
grey matter density.  From my SPM analysis I have 4 clusters in the  
brain that positively correlated with the creativity measures (p<. 
001, threshold of 25) and 4 clusters that negatively correlated (p<. 
001, threshold of 25) with the same creativity measures.  I then used  
Marsbar (love the name) to extract ROI (full options) and export the  
mean voxel value (summary time course for region) for each cluster of  
each scan.

I wanted the mean voxel value for each scan (rather than the peak  
voxel value that SPM gives) to make a scatter plot with the  
creativity scores and calculate correlation coefficients.  When I do  
this for my clusters that correlated positively with the creativity  
scores I get really nice correlations (between 0.7 to 0.62).   
However, when I plot the mean voxel values for the clusters that  
correlated negatively, i get really weak correlation coefficients (-. 
2  to 0.09).  I am wondering how this is possible to get such weak  
correlation coefficients for the negative correlations by just using  
the mean of the clusters, when in SPM I am getting these 4 clusters  
as significantly negatively correlating.  I tried using the median  
instead of the mean in Marsbar, but this did not help.

Any thoughts or suggestions as to why this is happening would be  
greatly appreciated.

Merci beaucoup.


Julie