 |
 |
Lena, The field of fmri statistical analysis is a bit controversial, so you may get replies that disagree with mine, but here goes! I think the reviewer is concerned that you are selectively reporting the most active part of the cluster by limiting the analysis to the data surrounding the peak voxel. That introduces bias into the results. Lots of early fmri papers did this, and people are more careful about it now. Strictly speaking, decisions about analysis should be made before the data is acquired, i.e. the statistical threshold that will be used, and the regions of interest that will be analyzed. Two methods that avoid bias are: 1) choosing the spherical region (or regions) for analysis before you look at the data, based on previous publications. 2) run two identical tasks on each patient, and use one of the datasets, at a chosen statistical threshold, to generate masks that are then used to limit the analysis of the second set of data. The data-generated mask can also be combined with a previously-chosen anatomic mask so that you analyze the most active part of the chosen anatomic mask. This way the experiment itself limits the region of interest, so there is no operator bias. Hope this helps! Jim On Sat, Mar 23, 2024 at 9:21 AM Lena Lim < [log in to unmask]> wrote: > Dear SPM Experts, > > Is it okay to extract beta values of sig clusters/regions using MarsBaR, > defined using spherical masks with a radius of 8mm around the peak > coordinates? I have seen many earlier papers extracting beta values using > spherical masks around peak coordinates too, but a reviewer commented > that a sphere centered at the peak coordinate is not representative of the > whole cluster, and that I should use the whole SPM clusters in their > original form instead... could you please kindly advise if the latter > approach is the better one please? > > > > Many thanks, > > > Lena > > >