> rather than
abs >u and z for activtions, rather than
gamma?
estimate_template * add custom smoothing * add parallel
over subjects
Fixes and cleaning up
Data pre-processing, for both template estimation and TemplateICA *
norm_BOLD replaces scale_BOLD, since now it
detrends the data too. * Do not scale across space by default. Instead,
scale across space only where necessary (one of the regressions in dual
regression, and PCA-based operations). * Option to detrend. * Compute
scale after centering and detrending.
Dual regression * Option to normalize A matrix
Template estimation * Merge estimate_template.cifti and
estimate_template.nifti into estimate_template
* Option to denoise each scan before computing DR * If using pseudo
retest data, split each subject’s scan after denoising and detrending,
not before. * Option for different variance template (non-negative) *
Option to obtain DR results * Return parameters, including data
normalization choices. * print and summary for
templates
TemplateICA * Merge templateICA.cifti and
templateICA.nifti into templateICA * Option to
return DR results in estimate_template * Option to skip
dimension reduction * Option to use parallel computing for iterating
over voxels. * Option to provide "xifti" or
"nifti" objects directly * Option to provide multiple scans
* For denoising: replace maxQ with Q2_max to
simplify logic * Fix cases where time_inds and
resamp_res are provided. * Return parameters, including
data normalization choices. * Replace subject IC variance with the
standard error * summary and plot for
activations
Misc * Implement sign matching * Refine group_ICA *
Refine make_mesh * Remove INLA required version (bring
back?)
For developers * Laid some groundwork for estimating templates using
>2 scans per subject: var_decomp can handle >2 scans
per subject (but need to test it!) * New code template calculation
Update functions that work with "xifti" objects to
conserve time and memory.