This program is the MUCH MUCH MUCH faster version of the original
3dSpaceTimeCorr program. It uses pre-Z-score-ization (like InstaCorr)
and OpenMP-based multithreading to achieve such relativistically better
runtimes.
This program is for calculating something *similar* to the (Pearson)
correlation coefficient between corresponding voxels between two data
sets, which is what 3dTcorrelate does. However, this is program
operates differently. Here, two data sets are loaded in, and for each
voxel in the brain:
+ for each data set, an ijk-th voxel is used as a seed to generate a
correlation map within a user-defined mask (e.g., whole brain,
excluding the seed location where r==1, by definition);
+ that correlation map is Fisher Z transformed;
+ the Z-correlation maps are (Pearson) correlated with each other,
generating a single correlation coefficient;
+ the correlation coefficient is stored at the same ijk-th voxel
location in the output data set;
and the process is repeated. Thus, the output is a whole brain map
of r-correlation coefficients for corresponding voxels from the two data
sets, generated by temporal and spatial patterns (-> space+time
correlation!).
This could be useful when researchers *wish* that they could use
3dTcorrelate on something like resting state FMRI data. Maybe.
Note that this program could take several minutes or more to run,
depending on the size of the data set and mask.
written by: PA Taylor (SSCC, NIMH, NIH, USA)
with a great deal of input, helpful discussion, beta-testing
and patience from Peter Kirk (EDB, NIMH, NIH, USA), who also
has the distinction of finding a real _use_ for this program.
