Re: [HCP-Users] How to threshold the task data?

[Harms, Michael]

It was created by submitting the results for all 440 subjects into a "Level 3" group-level analysis.  Basically the same as one would be doing in a group analysis in SPM or FSL, but applied to the CIFTI data.  If you aren't familiar with this process, I'd suggest that you review the FSL FEAT documentation on "group statistics".  See in particular the section on "single-group average": http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FEAT/UserGuide#Single-Group_Average_.28One-Sample_T-Test.29 In essence, for each grayordinate (and each provided task contrast), we just do a standard one-sample t-test using the 440 related subjects as input. cheers, -MH --  Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO  63110 Email: mha...@wustl.edu From: Long <longspac...@gmail.com> Date: Wednesday, November 26, 2014 3:12 PM To: "Harms, Michael" <mha...@wustl.edu> Cc: "hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] How to threshold the task data? Dear Michael, Thank you for the explaination. The maximum value is around 30 in one averaged motor task map across 440 subjects, but if 30 over 440 is around 0.07, then this value is a bit of low as the individual-level. How was the averaged activation map created? average across individual activation maps or average fMRI signal first then create the activation maps? Best, Xiangyu On Wed, Nov 26, 2014 at 6:04 PM, Harms, Michael <mha...@wustl.edu> wrote: Yes, there are going to be some large z-stats in a analysis involving 440 subjects! Currently, we don't have a tool yet for running a cluster level correction on the surfaces (or even more specifically, the full CIFTI grayordinate space, including both surface vertices and subcortical voxels -- there are some issues to work out on how to do that appropriately on the full CIFTI).  The basic idea is going to be to use 'randomise' to establish the null-distribution of either cluster size or a TFCE statistic.  But we are just in the early stages of working on that. cheers, -MH --  Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO  63110 Email: mha...@wustl.edu From: Long <longspac...@gmail.com> Date: Wednesday, November 26, 2014 10:55 AM To: "Harms, Michael" <mha...@wustl.edu> Cc: "hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] How to threshold the task data? Dear Michael, Thank you for your reply! I loaded right hand motor task data into Matlab and see the range of z score is from ~-20 to ~29. If I convert those high z scores, the p value will be just 0. Is that normal? also how can I run a cluster-level correction on the surface map? Best, Xiangyu On Wed, Nov 26, 2014 at 3:42 PM, Harms, Michael <mha...@wustl.edu> wrote: Hi, The file has "zstat" in its name, so it is the z-statistic from the Level 3 group model across 440 related subjects ("R440").  You could compute a p-value from a standard normal distribution, but those p-values will not be fully accurate, because as we note in the data release manual, the model for computing the z-stats didn't account for the familial relationships between subjects (i.e., the R440 subjects are not an independent group of subjects).  You could use the "U100" set of subjects instead, which are 100 unrelated subjects.  In that case, p-values computed from the z-stat should be accurate, but note that those will be grayordinate-wise p-values, and will not be corrected for multiple comparisons. cheers, -MH --  Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO  63110 Email: mha...@wustl.edu From: Long <longspac...@gmail.com> Date: Wednesday, November 26, 2014 6:50 AM To: "hcp-users@humanconnectome.org" <hcp-users@humanconnectome.org> Subject: [HCP-Users] How to threshold the task data? Dear all, I tried to run statistical analysis on the task data, for instance, HCP_Q1-Q6_R440_tfMRI_ALLTASKS_level3_zstat1_hp200_s2.dscalar.nii. However I don't know if the data contained z score or beta score, and how to calculate the p value from the task data. Thank you in advance for helping! Best, Xiangyu _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users   The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.   The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.   The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users