Hi,
I am using the slow_surf_clustsim.py to estimate the cluster size for multiple comparison.
One thing I found strange was that on one hemisphere of a subject, during a step that the script seem to be doing iterative smoothing until the targeted FWHM level, the FWHM actually decreases after it reaches a peak. I have not fully understood this iterative process. But I assume it is applying the same filtering kernel again and again until the estimated smoothness reaches the desired value. I do not quite understand how can it get less smooth after more filtering.
Here is my command of slow_surf_clustsim.py:
slow_surf_clustsim.py -save_script surf.clustsimu.$hemi.tcsh -uvar spec_file ../freesurfer/SUMA/std.$ld_num.${subj}_${hemi}.spec -uvar surf_vol ${subj}_SurfVol_Al+orig -uvar blur $target_FWHM -uvar results_dir clust.result.surface.$hemi -uvar pthr_list 0.05 0.01 0.002 -on_surface yes -uvar niter 2000
($hemi would be either lh or rh, the $ld_num was 141 in this case. And I set the $target_FWHM as 5.0 )
And the output looks like below (I truncated the output but clearly you can see the fwhm starts decreasing after 160 iterations). Any advice?
Thanks!
Mingbo
++ writing script to surf.clustsimu.rh.tcsh
have p-values: 0.05 0.01 0.002
have z-scores: 1.644854 2.326348 2.878162
Notice SUMA_PopulateDsetNodeIndexNel:
Assuming node indexing
is explicit.
1st row is for node 0
2nd is for node 1, etc.
++ Notice ConvertDset (SUMA_ConvertDset.c:881 @12:10:36):
Padding output dset until node 198811
++ 3dcalc: AFNI version=AFNI_2011_12_21_1014 (Oct 29 2014) [64-bit]
++ Authored by: A cast of thousands
*+ WARNING: input 'a' is not used in the expression
++ Output dataset ./all_1.gii
== iter block 001 (size 10) @ Wed Dec 3 12:10:36 CST 2014
++ 3dcalc: AFNI version=AFNI_2011_12_21_1014 (Oct 29 2014) [64-bit]
++ Authored by: A cast of thousands
++ Calculating 3D+time[10] dataset from 3D datasets and time series, with dt=1 s
*+ WARNING: input 'a' is not used in the expression
*+ WARNING: input 'b' is not used in the expression
++ Output dataset ././surf.noise.001.gii
1.14user 0.06system 0:01.42elapsed 84%CPU (0avgtext+0avgdata 19648maxresident)k
112inputs+0outputs (0major+5958minor)pagefaults 0swaps
Notice SUMA_PopulateDsetNodeIndexNel:
Assuming node indexing
is explicit.
1st row is for node 0
2nd is for node 1, etc.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.369328)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.444366)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.457216)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.465486)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.446360)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.342561)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.486983)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.352647)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.440036)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.495293)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_SigForFWHM (SUMA_GeomComp.c:7334 @12:10:40):
Low Sigma/AvgLe, increased it to 0.500
Expected niter now: 60
Kernel Bandwidth / Average Edge Distance = 0.500000/0.804949 = 0.402475
Corresponding Kernel Numerator = 0.135335
++ Notice SurfSmooth (SUMA_SurfSmooth.c:2217 @12:10:40):
Smoothing kernel bandwidth (sigma) = 0.402475
oo Warning SUMA_Chung_Smooth_Weights_07 (SUMA_GeomComp.c:3475):
Weights sum < 1.0f+1e-8 at node 48136
Mesh may be too coarse for kernel
bandwidth of 0.402475 in float precision.
Consider decreasing your number of iterations.
Future similar warnings are muted, but
a count is issued at the end.
oo Warning SUMA_Chung_Smooth_Weights_07 (SUMA_GeomComp.c:3500):
6 precision warnings out of 198812 nodes forming surface (0.00302 %).
Nodes with possible precision problems:
48136, 48199, 48261, 48376, 48434, 48435
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.369328)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.444366)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.457216)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.465486)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.446360)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.342561)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.486983)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.352647)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.440036)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
++ Notice SUMA_estimate_FWHM_1dif (SUMA_GeomComp.c:8923 @12:10:40):
Distribution of data is possibly random noise (p=0.495293)
Expect fwhm to be no different from 0
FWHM values up to 0.55(segments) or 0.44(mm)
are likely meaningless (at p=0.01) on this mesh.
Iteration 0, fwhm = 0.199925; target 5.000000
Iteration 10, fwhm = 1.490346; target 5.000000
Iteration 20, fwhm = 1.725553; target 5.000000
Iteration 30, fwhm = 1.850237; target 5.000000
Iteration 40, fwhm = 1.928905; target 5.000000
Iteration 50, fwhm = 1.982976; target 5.000000
Iteration 60, fwhm = 2.022015; target 5.000000
Iteration 70, fwhm = 2.051033; target 5.000000
Iteration 80, fwhm = 2.072949; target 5.000000
Iteration 90, fwhm = 2.089603; target 5.000000
Iteration 100, fwhm = 2.102226; target 5.000000
Iteration 110, fwhm = 2.111686; target 5.000000
Iteration 120, fwhm = 2.118613; target 5.000000
Iteration 130, fwhm = 2.123484; target 5.000000
Iteration 140, fwhm = 2.126664; target 5.000000
Iteration 150, fwhm = 2.128442; target 5.000000
Iteration 160, fwhm = 2.129047; target 5.000000
Iteration 170, fwhm = 2.128664; target 5.000000
Iteration 180, fwhm = 2.127447; target 5.000000
Iteration 190, fwhm = 2.125520; target 5.000000
Iteration 200, fwhm = 2.122988; target 5.000000
Iteration 210, fwhm = 2.119939; target 5.000000
Iteration 220, fwhm = 2.116447; target 5.000000
Iteration 230, fwhm = 2.112575; target 5.000000
Iteration 240, fwhm = 2.108376; target 5.000000
Iteration 250, fwhm = 2.103896; target 5.000000
Iteration 260, fwhm = 2.099175; target 5.000000
Iteration 270, fwhm = 2.094247; target 5.000000
Iteration 280, fwhm = 2.089141; target 5.000000
Iteration 290, fwhm = 2.083884; target 5.000000
Iteration 300, fwhm = 2.078499; target 5.000000
Iteration 310, fwhm = 2.073004; target 5.000000
Iteration 320, fwhm = 2.067418; target 5.000000
Iteration 330, fwhm = 2.061755; target 5.000000
Iteration 340, fwhm = 2.056031; target 5.000000
Iteration 350, fwhm = 2.050255; target 5.000000
Iteration 360, fwhm = 2.044440; target 5.000000
Iteration 370, fwhm = 2.038595; target 5.000000
Iteration 380, fwhm = 2.032728; target 5.000000
Iteration 390, fwhm = 2.026847; target 5.000000
Iteration 400, fwhm = 2.020958; target 5.000000
Iteration 410, fwhm = 2.015068; target 5.000000
Iteration 420, fwhm = 2.009181; target 5.000000
Iteration 430, fwhm = 2.003303; target 5.000000
Iteration 440, fwhm = 1.997438; target 5.000000
Iteration 450, fwhm = 1.991590; target 5.000000
Iteration 460, fwhm = 1.985761; target 5.000000
Iteration 470, fwhm = 1.979956; target 5.000000
Iteration 480, fwhm = 1.974177; target 5.000000
Iteration 490, fwhm = 1.968426; target 5.000000
Iteration 500, fwhm = 1.962706; target 5.000000
Iteration 510, fwhm = 1.957017; target 5.000000
Iteration 520, fwhm = 1.951363; target 5.000000
Iteration 530, fwhm = 1.945744; target 5.000000
Iteration 540, fwhm = 1.940162; target 5.000000
Iteration 550, fwhm = 1.934618; target 5.000000
Iteration 560, fwhm = 1.929112; target 5.000000
Iteration 570, fwhm = 1.923646; target 5.000000
Iteration 580, fwhm = 1.918220; target 5.000000
Iteration 590, fwhm = 1.912834; target 5.000000
Iteration 600, fwhm = 1.907491; target 5.000000
Iteration 610, fwhm = 1.902188; target 5.000000
Iteration 620, fwhm = 1.896928; target 5.000000