Subject: | | Re: Reporting ROI analyses and correcting for multiple comparisons |
From: | | Roberto Viviani <[log in to unmask]> |
Reply-To: | | [log in to unmask][log in to unmask]> wrote:
>Dear all, > >I get the following error message when running SPM8, with example data >sets. I don't seem to be able to run an analysis that used to be >straightforward. > >Best, >Cedric > >----------------------------------------------------------------------- >Running job #1 >----------------------------------------------------------------------- >Running 'Realign: Estimate & Reslice' >Failed 'Realign: Estimate & Reslice' >Error using ==> class >Cannot change the number of fields of class 'file_array' without first >typing 'clear classes'. >In file "[log in to unmask]" (v1340), >function "file_array" at line 41. >In file "[log in to unmask]" (v1143), function >"nifti" at line 70. >In file "/home/cgineste/sft/spm8/spm8/spm_vol_nifti.m" (v1143), function >"spm_vol_nifti" at line 16. >In file "/home/cgineste/sft/spm8/spm8/spm_vol.m" (v2237), function >"subfunc" at line 138. >In file "/home/cgineste/sft/spm8/spm8/spm_vol.m" (v2237), function >"subfunc1" at line 84. >In file "/home/cgineste/sft/spm8/spm8/spm_vol.m" (v2237), function >"subfunc2" at line 70. >In file "/home/cgineste/sft/spm8/spm8/spm_vol.m" (v2237), function >"spm_vol" at line 54. >In file "/home/cgineste/sft/spm8/spm8/spm_realign.m" (v1265), function >"spm_realign" at line 101. >In file "/home/cgineste/sft/spm8/spm8/config/spm_run_realign_estwrite.m" >(v1185), function "spm_run_realign_estwrite" at line 26. > >The following modules did not run: >Failed: Realign: Estimate & Reslice > > > >-- >Cedric Ginestet >Centre for Neuroimaging Sciences >NIHR Biomedical Research Centre >Institute of Psychiatry, Box P089 >Kings College London >De Crespigny Park >London >SE5 8AF >Tel: (+44) 79-2051-3529 >Fax: (+44) 20-7402-2150 >email1: [log in to unmask] >email2: [log in to unmask] >47_4Dec200921:18:[log in to unmask] |
Date: | | Fri, 11 Dec 2009 11:24:35 +0100 |
Content-Type: | | text/plain |
Parts/Attachments: |
|
|
|
|
>> It seems that a majority of papers publishing ROI analyses does not apply
>> correction for multiple comparisons.
>>
>> The rationale behind it, I suppose, being that ROI analyses correspond to
>> planned comparisons and therefore are exempt from usual corrections applied
>> to post-hoc tests.
>>
>> - Is there a relevant reference that supports that point ?
The point of doing a ROI analysis, when a parametric map analysis
would be available, is precisely that of avoiding to correct for the
number of voxels. You still should apply correction for the number of
comparisons, as Stephen notes, but the correction is more lenient
because you'll have a few ROIs not n thousand voxels. There is no
reference on this specifically, it is just follows logically from what
frequentist significance values are intended to be; all repeated tests
must be corrected. But you might want to have a look at the Intro
chapter in Hochberg and Tamhane for definitions of familiywise
corrections, experimentwise corrections etc.
>> - Can this logic also be applied to ROIs that are dependent of the dataset
>> tested ie when ROIs have been generated using main effects of a factorial
>> design rather than using a functional localiser ?
It very much depends on how your 'factorial design' is defined. In an
event-related design, the regressors may possess some significant
degree of collinearity after convolution with the BOLD canonical
response. Kriegeskorte et al. in a recent paper in Nature Neuroscience
raised the issue of the role of the weight matrix modelling the
temporal correlation of epi data. Importantly, orthogonal contrast
weights do not automatically translate into orthogonal contrasts if
the underlying regressors are not the orthogonal ANOVA dummies. You
have to think about how much of the ANOVA logic can be applied when
your design is no longer an ANOVA.
If you want to have a look at a way to make an event-related design
factorial-like, see Elliott et al., NeuroImage 21:984-990.
Best wishes,
Roberto Viviani
University of Ulm, Germany
|
|
|
|