Dear Satoru,
thanks for your response! Please see my comments below.
>> In order to avoid the known problem with too large size of SnPM_ST.mat
>> I tried to threshold the cluster-defining pseudo t value at the
>> setup stage. However, in this case SnPM does not create the
>> SnPM_ST.mat and does not allow me to get voxel-cluster combined
> results.
>> If I do not use the threshold, SnPM_ST.mat (about 2GB) is created.
>
> Unfortunately voxel-cluster combining, as currently implemented,
> requires the SnPM_ST.mat file. Voxel-cluster combining needs to do the
> permutation test twice, the first permutation test to calculate
> corrected voxel and cluster p-values, then the second permutation test
> to calculate p-values for voxel-cluster combined results. By using the
> SnPM_ST.mat file, SnPM can do these two permutation tests without
> repeating the "Compute" step twice.
>
> Well, the large SnPM_ST file is a known SnPM problem unfortunately. By
> selecting the cluster-defining threshold at the SetUp step, were you be
> able to find any voxels or clusters surviving FWE p<0.05? The reason I
> am asking this is because when there is no activation, the SnPM_ST file
> tends to be large when the peudo-t statistic is used.
I was simply wondering why when I choose to collect supra-threshold
stats and I do not enter a threshold at the setup step, everything is
completed OK (and, yes, there are both voxels and clusters surviving
FWE p<0.05) and I can see results (including "combo"), change
thresholds at the "result" step etc. However, if I try to select a
threshold at the setup stage, keeping all other parameters same, then,
at the "voxel-cluster results" step I get this error in the matlab window:
=======================================================================
??? Error using ==> snpm_combo_pp
Suprathreshold stats not collected! Cannot do cluster-combining!
??? Error while evaluating uicontrol Callback
-----------------------------------------------------------------------
... and I notice that SnPM_ST.mat file was not created.
Just in case, here is an example log:
--------------------------------------------
spm_input : Select design type...: (MultiSub: One Sample T test on
differences; 1 condition) 3
spm_get : Select all scans :
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat03_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat04_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat06_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat08_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat09_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat10_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat11_con_0010.img
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\pat14_con_0010.img
spm_input : # of confounding covariates: (0) 0
spm_input : 256 Perms. Use approx. test?: (yes) y
spm_input : # perms. to use? (Max 256): (256) 256
spm_input : FWHM(mm) for Variance smooth: (10) 10
spm_input : Collect Supra-Threshold stats?: (yes) 1
spm_input : Define the thresh now?: (yes) 1
spm_input : Clus-def thresh (pseudo t): (4) 4
spm_input : Select global normalisation: (<no global normalisation>) 1
spm_input : grand mean scaling: (<no grand Mean scaling>) 2
spm_input : Threshold masking: (none) 1
----------------------------------------------------------------
THEN I TRY TO SEE RESULTS:
spm_get : Select SnPMcfg.mat CfgFile... :
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\SnPM_A\test_thresh\SnPMcfg.mat
spm_get : Select SnPM.mat for analysis... :
E:\MSCLER\PAT\CON_PAT_HP426_CORR\CON_10_SESS2_Task_Base\SnPM_A\test_thresh\SnPM.mat
spm_input : Positive or negative effects?: (+ve) 0
spm_input : Write filtered statistic img?: (no) 0
spm_input : Corrected p value for filtering: (0.05) 0.05
------------------------------------------------------------
AND AT THIS POINT I GET THE ERROR MESSAGE
It seems it does not matter which threshold I choose when answering
the question "Clus-def thresh (pseudo t)" at the setup step.
I get the error message anyway. This puzzles me...
Victor
PS: sorry for such a long message
|