Hi Olivier,

If you are referring to the beta_.nii images in the second-level results folders you are right that those reflect the raw/unstandardized regression coefficients from your GLM model (e.g. if you have two groups you should have two beta_.nii images, each representing the difference in connectivity between time1 and time2 in each of the two groups), and there is no simple way to standardize these effects (e.g. you could square these values and divide them by the residual intersubject variance to obtain eta-squared values for these effects using the information from the ResMS.nii and SPM.mat files in the same folder). On the other hand, perhaps you are referring to the effect sizes displayed in the barplots of your second-level analysis results, which show the effect-size associated with the Group-by-time interactions directly (these values represent the "difference of the differences" of the connectivity values within each cluster, e.g. (Group2Time2 - Group2Time1) - (Group1Time2 - Group1Time1)). If this is the case, you are also right that the effect-sizes reported here are also raw/unstandardized, but perhaps in this case you could more simply compute eta-square measures associated with these interaction effects by transforming the T-statistics reported in the same display (the associations for a 2x2 groups-by-time interaction will be eta-square = T-squared/(T-squared + dof))

Hope this helps

Alfonso

Originally posted by Olivier Brown:

Hi,

I ran a simple Group x Time interaction seed-to-voxel analysis and found a significant cluster. In the results explorer, I extracted effect sizes (beta coefficients) for reporting in the manuscript. However, it is my understanding that these are the unstandardized values. Is there a simple method for extracting the standardized effect sizes? 

Warm regards,

Olivier