One quote from Poldrack et al. before I hop off the soapbox: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2287206/ "Finally, while thresholds must account for the multiplicity of tests, we do encourage authors to make available unthresholded statistic and effect size images in order to display the whole range of effects in the data, including those that do not reach significance. These maps also make it easier to compare effect sizes across studies and increase the options for future meta-analyses."
On Oct 24, 2013, at 7:23 PM, Michael Cohen <[email protected]> wrote: > Thanks for the reply--this is an interesting suggestion, and it does seem > like one way to accomplish what I am trying to do. However, it also seems > like a fairly time-consuming solution, especially given that we're just > trying to get a good visualization of data that we've already analyzed > elsewhere. It's not -- especially if you script it, which I could help with if you want at some point. But I get that you want to get your results out sooner rather than later. > So, I'm just curious to see whether you (or anybody else on the list) might > have insight into the original questions that I had asked? Barring any > additional guidance, I think we may just use the interpolated voxel algorithm > on the cluster-thresholded FSL data, without any additional thresholding in > Caret. But I just would like to make sure that this is a reasonable > approach--since if there's a setting or two that we should tweak to get an > image that more accurately represents the data, it would be great to know > that before we submit these figures for publication. I think the only strict contraindication for interpolated voxel are cases analogous to where you'd use nearest-neighbor algorithms in volume-land (e.g., label/ROI/parcel volume). I think using interpolated voxel is fine, even with your thresholded image. Sure, you'll fade a bit at the edges, but if this is a concern, use enclosing voxel. Whether it's a concern depends on the nature and extent of the mapped data. In most cases, I doubt it will be a concern. My take anyway. > > Thanks, > Michael > _______________________________________________ > caret-users mailing list > [email protected] > http://brainvis.wustl.edu/mailman/listinfo/caret-users _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
