Hi,
In a group level RFX, when contrasting Group A>Group B, we find a specific region/cluster of difference. If we go to the peak voxel in that cluster, and then select PLOT>CONTRAST ESTIMATES and select the specific contrast, we get a single bar plot, and when we type "beta" in the Matlab window, it provides us with the 3 values - one mean parameter estimate (beta value) for GroupA, one for GroupB and the sum of the two (although I'm clueless on how to get the standard error values (red bar) out of the plots in SPM2 - I think in SPM99 you could just type SE). Anyway, as we understand it, these values are extracted from the beta*.img files in the group level RFX directory. But we've also been using a beta extracting script from Ling Wang (<http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0605&L=SPM&P=R66437&I=-3>), which goes back one step and instead extracts the parameter estimate values from each individual subject's beta*.img files, per contrast at this specified peak voxel location. Are we to expect that when we have all these values across subjects, that the average across subjects should match/equate to the beta value we get from the plot at the RFX level? Right now it is not even close. We're just trying to figure out how to get the best measure of these beta values, working back from our regions of significant difference at the RFX level.
Many thanks for any help anyone can give,
-Peter Hu
|