AFNI program: 3dttest

Gosset (Student) t-test sets of 3D datasets
Usage 1: 3dttest [options] -set1 datasets ... -set2 datasets ...
   for comparing the means of 2 sets of datasets (voxel by voxel).

Usage 2: 3dttest [options] -base1 bval -set2 datasets ...
   for comparing the mean of 1 set of datasets against a constant.

OUTPUTS:
 A single dataset is created that is the voxel-by-voxel difference
 of the mean of set2 minus the mean of set1 (or minus 'bval').
 The output dataset will be of the intensity+Ttest ('fitt') type.
 The t-statistic at each voxel can be used as an interactive
 thresholding tool in AFNI.

t-TESTING OPTIONS:
  -set1 datasets ... = Specifies the collection of datasets to put into
                         the first set. The mean of set1 will be tested
                         with a 2-sample t-test against the mean of set2.
                   N.B.: -set1 and -base1 are mutually exclusive!
  -base1 bval        = 'bval' is a numerical value that the mean of set2
                         will be tested against with a 1-sample t-test.
  -set2 datasets ... = Specifies the collection of datasets to put into
                         the second set.  There must be at least 2 datasets
                         in each of set1 (if used) and set2.
  -paired            = Specifies the use of a paired-sample t-test to
                         compare set1 and set2.  If this option is used,
                         set1 and set2 must have the same cardinality.
                   N.B.: A paired test is intended for use when the set1 and set2
                         dataset function values may be pairwise correlated.
                         If they are in fact uncorrelated, this test has less
                         statistical 'power' than the unpaired (default) t-test.
                         This loss of power is the price that is paid for
                         insurance against pairwise correlations.
  -unpooled          = Specifies that the variance estimates for set1 and
                         set2 be computed separately (not pooled together).
                         This only makes sense if -paired is NOT given.
                   N.B.: If this option is used, the number of degrees
                         of freedom per voxel is a variable, rather
                         than a constant.
  -dof_prefix ddd    = If '-unpooled' is also used, then a dataset with
                         prefix 'ddd' will be created that contains the
                         degrees of freedom (DOF) in each voxel.
                         You can convert the t-value in the -prefix
                         dataset to a z-score using the -dof_prefix dataset
                         using commands like so:
           3dcalc -a 'pname+orig[1]' -b ddd+orig \
                  -datum float -prefix ddd_zz -expr 'fitt_t2z(a,b)'
           3drefit -substatpar 0 fizt ddd_zz+orig
                         At present, AFNI is incapable of directly dealing
                         with datasets whose DOF parameter varies between
                         voxels.  Converting to a z-score (with no parameters)
                         is one way of getting around this difficulty.
  -workmem mega      = 'mega' specifies the number of megabytes of RAM
                         to use for statistical workspace.  It defaults
                         to 12.  The program will run faster if this is
                         larger (see the NOTES section below).

The -base1 or -set1 command line switches must follow all other options
(including those described below) except for the -set2 switch.

INPUT EDITING OPTIONS: The same as are available in 3dmerge.

OUTPUT OPTIONS: these options control the output files.
  -session  dirname  = Write output into given directory (default=./)
  -prefix   pname    = Use 'pname' for the output directory prefix
                       (default=tdif)
  -datum    type     = Use 'type' to store the output difference
                       in the means; 'type' may be short or float.
                       How the default is determined is described
                       in the notes below.

NOTES:
 ** To economize on memory, 3dttest makes multiple passes through
      the input datasets.  On each pass, the entire editing process
      will be carried out again.  For efficiency's sake, it is
      better to carry out the editing using 3dmerge to produce
      temporary datasets, and then run 3dttest on them.  This applies
      with particular force if a 'blurring' option is used.
      Note also that editing a dataset requires that it be read into
      memory in its entirety (so that the disk file is not altered).
      This will increase the memory needs of the program far beyond
      the level set by the -workmem option.
 ** The input datasets are specified by their .HEAD files,
      but their .BRIK files must exist also! This program cannot
      'warp-on-demand' from other datasets.
 ** This program cannot deal with time-dependent or complex-valued datasets!
      By default, the output dataset function values will be shorts if the
      first input dataset is byte- or short-valued; otherwise they will be
      floats.  This behavior may be overridden using the -datum option.

INPUT DATASET NAMES
-------------------
This program accepts datasets that are modified on input according to the
following schemes:
  'r1+orig[3..5]'                                    {sub-brick selector}
  'r1+orig<100.200>'                                 {sub-range selector}
  'r1+orig[3..5]<100..200>'                          {both selectors}
  '3dcalc( -a r1+orig -b r2+orig -expr 0.5*(a+b) )'  {calculation}
For the gruesome details, see the output of 'afni -help'.