 |
 |
Hi - I think I hear Tom weeping, 40 miles away in Warwick…. On 7 Jul 2011, at 18:42, Alisha Janssen wrote: > Hello FSL experts, > > I have finished a recent analysis of two groups (one clinical, one healthy) using dual_regression. Because of the stringent thresholds used in randomise This premise is incorrect I'm afraid. Permutation testing will give you accurate p-values even if your data and processing may have introduced non-Gaussianities (etc) into your residuals, which may well be the case here, hence why you should use it for the output of dual-regression. Permutation testing is more accurate (or the same in the ideal world for both) than Gaussian random field theory even if your residuals are Gaussian. You should not use the latter in the hope that it will be less stringent! Likewise, getting FWE (family wise error) corrected p-values from permutation testing is not over-stringent in terms of how the FWE is estimated - it is correct. What you *might* decide is too "stringent" is to use FWE rather than FDR - but that decision is not about whether you use permutation testing or not. If you want to use FDR, then take uncorrected p-values from randomise into the fdr program - see the manual for that. You can also possibly raise sensitivity by using the TFCE option in randomise, if you haven't already done that. Hope that makes sense, Cheers Steve > , I decided to rethreshold the data using a different method. According to previous emails to the forum I have seen, I applied the following two commands to my raw dr_stage2_tstat images; > > fslmaths dr_stage3_ic*_tstat1.nii.gz -mul 0 -add 1 grot.nii > ttoz grot.nii dr_stage3_ic*_tstat1.nii.gz 74 (based on 79 participants with 5 E.V.'s in my glm). > > I then used easythresh on these z-stat images to rethreshold. Now I have cluster masks for each of my components of interest. My next step was to pull out the fisher-transformed z-values for each individual to test the effect of clinical variables (disease duration, etc.). With previous data, I have used the featquery command to calculate the mean z-value of a cluster in each individual's data. > > In this case, I will need to extract these values from the dr_stage2 outputs for each individual. I was wondering what command I could use that is similar to featquery, but doesn't require a feat directory to extract the z-values of my clusters from each individual's stage2 data? Or is this possible? > --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Associate Director, Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 222726 (fax 222717) [log in to unmask] http://www.fmrib.ox.ac.uk/~steve ---------------------------------------------------------------------------