3dXClustSim


This program takes as input random field simulations
(e.g., from 3dttest++) and does the ETAC processing to
find cluster figure of merit (FOM) thresholds that are
equitable (AKA balanced) across
  * voxel-wise p-values (-pthr option)
  * blurring cases      (-ncase option)
  * H power values      (-hpow option) -- probably not useful
as well as being balanced across space to produce
a False Positive Rate (FPR) that is approximately the
same for each location and for each sub-case listed
above. The usual goal is a global FPR of 5%.

* This program can be slow and consume a LOT of memory!
  (And I mean a BIG LOT, not a small lot.)

* The output is a set of multi-threshold (*.mthresh.*.nii)
  files -- one for each of the -ncase inputs.

* These files can be used via program 3dMultiThresh
  to produce an 'activation' mask.

* 3dXClustSim is intended to be used from 3dttest++
  (via its '-ETAC' option) or some other script.

* It is not intended to be run directly by any but the most
  knowledgeable and astute users. Which is why this help is so terse.

--------
OPTIONS:
--------

 -inset       mask sdata ... {MANDATORY} [from 3dtoXdataset or 3dttest++]
 -insdat                     Data files are in the '.sdat' format.

 -NN          1 or 2 or 3    [-NN1 or -NN2 or -NN3 will work; default = 2]
 -sid         1 or 2         [-1sid or -2sid will work; default = 2]
 -hpow        0 1 2          [or some subset of these; default = 2]

 -ncase       N lab1 .. labN [multiple processing cases; e.g., blurs]
                             [default = 1 A]
                             [example = 4 b04 b06 b08 b10]

 -pthr        list of values [default = 0.0100 0.0056 0.0031 0.0018 0.0010]
                             [equiv z1= 2.326  2.536  2.731  2.911  3.090 ]
                             [equiv z2= 2.576  2.770  2.958  3.121  3.291 ]

 -FPR ff      set global FPR goal to ff%, where ff is an integer
              from 2 to 9 (inclusive). Default value is 5.

 -multiFPR    compute results for multiple FPR goals (2%, 3%, ... 9%)

 -minclust M  don't allow clusters smaller than M voxels [default M=5]

 -local       do the 'local' (voxelwise) ETAC computations
 -global      do the 'global' (volumewise) ETAC computations
 -nolocal     don't do the 'local'
 -noglobal    don't do the 'global'

 -splitfrac F split simulations into pieces ( 0.2 < F < 0.8 )

 -prefix      something useful
 -verb        be more verbose
 -quiet       silentium est aureum

**-----------------------------------------------------------
** Authored by Lamont Cranston, also known as ... The Shadow.
**-----------------------------------------------------------