 |
 |
Hi Dang, If you are testing the intercept, then you need to include the option -ise, otherwise there is nothing to permute... All the best, Anderson On Sun, 18 Nov 2018 at 20:01, Dang Zheng <[log in to unmask]> wrote: > Dear Dr. Winkler, > > Recently, I have read one of your papers published on NeuroImage (2015), > “Multi-level block permutation”. I think this strategy is very useful for > me to address the problem of controlling the non-independence of > observation within twin pairs in our twin study, and we hope to used it in > our data analysis. I have tried with palm of fsl, but it seems that the > result is really strange. Before using palm, we used spm12 to conduct one > sample t-test to detect the significantly positive and negative functional > connectivity (FC) of a group of 188 subjects (94 pairs of twins), and we > found significant FC in many brain regions (voxel level FWE corrected > p<0.05, see attachment Fig1). However, after using palm, none voxel > survived. Even though the permutation test is a strict correction, the > blank result is hard for me to believe. So I am writing to ask for your > advice to find out which one of my follow steps is incorrect: > 1. I used fslmerge to merge the 188 subjects’ FC maps to a 4D.nii image. > 2. The t-test design is as follow: EV1 is the FC images, and EV2-EV4 are > three control variables (age, gender and headmotion, see attachment Fig2) > 3. The definition of exchangeability blocks is blocks.csv (please see the > attachment). In our dataset, all the twin families have only two twin > children (a pair of MZ or DZ twins) without other non-twin siblings. We > adopt the design in your paper’s (“Multi-level block permutation”) Fig.7. > The second column represents that there are two kinds of family (1= MZ > family, 2= DZ family), the third column represents their family ID, and the > forth column represents twins’ zygnosity (1= MZ, 2= DZ). The last column > represents the 188 subjects’ ID. > 4. Run in matlab: -i 4D.nii -d meta_FC.mat -t meta_FC.con -eb > blocks_test.csv -T -n 500 -o test > 5. Viewing the statistic image (palm_vox_tstat.nii), it looks similar to > the previous statistic image got by spm12. But in the p-value image > (test_tfce_tstat_fwep.nii or even test_vox_tstat_uncp.nii), it seems all > the voxels’ p value is 1. > (I attached my data and results for your reference.) > > Thank you very much, and I am looking forward to your answers and > suggestions. > > Best wishes, > Dang > > ######################################################################## > > To unsubscribe from the FSL list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1